Autonomous aerial robots must operate in cluttered, wind-disturbed environments where turbulence and gusts generated by wind-object and terrain interactions introduce significant aerodynamic risks, including orientation instability, sensor degradation, control drift, and increased power consumption, often leading to mission failure or crash. We present Graphlets-based Zero-Shot Planning Framework (GZS), a novel, non-parametric, fast computation, memory-efficient, zero-shot training-free onboard inference framework for real-time 3D spatial-aware aerodynamic risk perception that operates without prior scene knowledge. GZS dynamically classifies point clouds to extract local topology, incorporates physics-informed modeling of wind interactions, and applies attention-guided segment matching to generate onboard 3D representations of wind-induced aerodynamic risk. It transforms unstructured scene segments into structured graphlets topologies encoding aerodynamic risk-aware features, enabling UAVs to identify and navigate through regions of minimal aerodynamic hazard in real time and without prior training in any environment. Unlike computational fluid dynamics(CFD)-based, deep learning, or map-dependent approaches, GZS performs zero-shot aerodynamic risk estimation in previously unseen and dynamic conditions. Extensive experiments demonstrate 90-95% accurate aerodynamic risk zone identification compared to conventional methods of CFDs and wind tunnels, while substantially reducing computational and memory overhead, and a 100% success rate in creating onboard 3d spatial-aware risk perceptions. Our results establish GZS as a framework for a zero-shot, non-parametric, robust, aerodynamic risk perception for autonomous real-time trajectory planning in wind-affected aerial environments.

Forecasting future sentiment offers valuable insight for high-impact domains such as advertising, recommender systems, and government decision-making, enabling better understanding and preparation for societal reactions. While most sentiment analysis methods focus on classifying the sentiment of already-expressed content, they remain inherently retrospective. In contrast, we address the underexplored challenge of forecasting user sentiment before it is publicly expressed. Specifically, we aim to predict how individuals will emotionally respond to real-world events based solely on their historical social media posts. We propose PRESTIGE (PRE-event Sentiment prediction Through Integrated Graph Embeddings), a framework that combines sentence-level semantic embeddings with topic-based user graph construction to capture user-level and topical context. Users are modeled as nodes enriched with embeddings of their pre-event posts and connected based on thematic similarity. Evaluation on three real-world Twitter (now X) datasets demonstrates the feasibility of pre-event sentiment prediction. We also find that deeper user histories significantly improve performance, suggesting that longer-term behavioral context plays a crucial role in accurate forecasting.

Landslides triggered by heavy rainfall threaten infrastructure and public safety, especially in regions with complex terrain and variable soils. Existing satellite-based methods often overlook the physics of landslide processes. We propose a ground-sensor-based framework that couples physics-informed machine learning with soil hydrology for improved landslide modeling. Our pedotransfer function (PTF)-enhanced physics-informed neural network estimates spatiotemporal soil moisture using Richards’ equation, operating under partially observed conditions without explicit boundary inputs, enabling scalable deployment. From modeled soil moisture, we derive physically consistent daily features to train a downstream landslide classifier. Across 43 sites in six US states, our approach achieves strong soil moisture prediction (R2= 0.77) and accurate landslide classification (Macro F1 = 0.88). These results highlight the value of integrating physics-based modeling with machine learning for practical geohazard prediction. Data and code are here.

Weather-related disruptions have a significant impact on a variety of industries, including agriculture, infrastructure, and public safety. Predicting these unusual weather events remains a significant challenge. The problem is complicated by the lack of high-quality weather data due to the failure of sensors at weather stations during severe weather. In this work, we proposed a novel method to classify rare severe weather-related events by incorporating publicly available tweets with the meteorological conditions readings collected from weather stations across Alaska. The use of multimodal data of varying quality is introduced to compensate for missing meteorological recordings obtained from the weather stations. In our study, we collected geotagged tweets from the region of focus and utilized context-aware BERT embeddings to rigorously analyze and ensure the validity and dependability of the social media texts. Labels for the social sensor data were generated based on weather events associated with the tweet collection. For predicting rare weather events, we proposed a multiclass classification model. This machine learning model was trained and tested using data from the year 2020. The results obtained by learning from the integrated data showed a significant increase in the F1 score when compared to relying on weather data alone. Our findings indicate that a model supplemented with daily weather and social media text data outperforms alternatives enhanced with hourly data. The proposed model achieved an F1-score of 0.83 for multimodal data, compared to 0.30 obtained by the baseline model that relies solely on weather data. Training the proposed model with the combined dataset significantly improved performance, resulting in a 95% accuracy.

This work investigates how the rich semantic embeddings of pre-trained language models can be used to help understand the general attitudes of an online community. This work describes a novel prediction model that can ingest statements describing an arbitrary context and a piece of content, and output answers to a set of 'attitude questions' describing the relationship between them. Typically, annotating answers to questions like "Does this contain sarcasm?", or "Is this content positive with respect to this context?" requires costly human interaction. In this work, we consider the ability of large language models to answer these questions, while under the constraint of a small dataset using a novel prediction head. We show that this methodology can accurately answer these attitude questions, compare the model to off-the-shelf language model approaches, and describe a method for collecting and annotating attitude question data sets. The novel attitude question answering model achieves a 89% accuracy on the attitude question answering task, outperforming the ablated models (87%) as well as the off the shelf models using BERT-based Sequence Classification (13%), BART-based Natural Language Inference (88%), and RoBERTa-based Question-Answering (87%).

MOTIVATION: Timetrees depict evolutionary relationships between species and the geological times of their divergence. Hundreds of research articles containing timetrees are published in scientific journals every year. The TimeTree (TT) project has been manually locating, curating and synthesizing timetrees from these articles for almost two decades into a TimeTree of Life, delivered through a unique, user-friendly web interface (timetree.org). The manual process of finding articles containing timetrees is becoming increasingly expensive and time-consuming. So, we have explored the effectiveness of text-mining approaches and developed optimizations to find research articles containing timetrees automatically. RESULTS: We have developed an optimized machine learning system to determine if a research article contains an evolutionary timetree appropriate for inclusion in the TT resource. We found that BERT classification fine-tuned on whole-text articles achieved an F1 score of 0.67, which we increased to 0.88 by text-mining article excerpts surrounding the mentioning of figures. The new method is implemented in the TimeTreeFinder (TTF) tool, which automatically processes millions of articles to discover timetree-containing articles. We estimate that the TTF tool would produce twice as many timetree-containing articles as those discovered manually, whose inclusion in the TT database would potentially double the knowledge accessible to a wider community. Manual inspection showed that the precision on out-of-distribution recently published articles is 87%. This automation will speed up the collection and curation of timetrees with much lower human and time costs. AVAILABILITY AND IMPLEMENTATION: https://github.com/marija-stanojevic/time-tree-classification. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

With the increase in volume of daily online news items, it is more and more difficult for readers to identify news articles relevant to their interests. Thus, effective recommendation systems are critical for an effective user news consumption experience. Existing news recommendation methods usually rely on the news click history to model user interest. However, there are other signals about user behaviors, such as user commenting activity, which have not been used before. We propose a recommendation algorithm that predicts articles a user may be interested in, given her historical sequential commenting behavior on news articles. We show that following this sequential user behavior the news recommendation problem falls into in the class of session-based recommendation. The techniques in this class seek to model users' sequential and temporal behaviors. While we seek to follow the general directions in this space, we face unique challenges specific to news in modeling temporal dynamics, e.g., users' interests shift over time, users comment irregularly on articles, and articles are perishable items with limited lifespans. We propose a recency-regularized neural attentive framework for session-based news recommendation. The proposed method is able to capture the temporal dynamics of both users and news articles, while maintaining interpretability. We design a lag-aware attention and a recency regularization to model the time effect of news articles and comments. We conduct extensive empirical studies on 3 real-world news datasets to demonstrate the effectiveness of our method.

Producing high-quality labeled data is a challenge in any supervised learning problem, where in many cases, human involvement is necessary to ensure the label quality. However, human annotations are not flawless, especially in the case of a challenging problem. In nontrivial problems, the high disagreement among annotators results in noisy labels, which affect the performance of any machine learning model. In this work, we consider three noise reduction strategies to improve the label quality in the Article-Comment Alignment Problem, where the main task is to classify article-comment pairs according to their relevancy level. The first considered labeling disagreement reduction strategy utilizes annotators’ background knowledge during the label aggregation step. The second strategy utilizes user disagreement during the training process. In the third and final strategy, we ask annotators to perform corrections and relabel the examples with noisy labels. We deploy these strategies and compare them to a resampling strategy for addressing the class imbalance, another common supervised learning challenge. These alternatives were evaluated on ACAP, a multiclass text pairs classification problem with highly imbalanced data, where one of the classes represents at most 15% of the dataset’s entire population. Our results provide evidence that considered strategies can reduce disagreement between annotators. However, data quality improvement is insufficient to enhance classification accuracy in the article-comment alignment problem, which exhibits a high-class imbalance. The model performance is enhanced for the same problem by addressing the imbalance issue with a weight loss-based class distribution resampling. We show that allowing the model to pay more attention to the minority class during the training process with the presence of noisy examples improves the test accuracy by 3%.

The methods for power system event detection using field-recorded data from Phasor Measurement Units (PMUs) often require many labeled events, which can be costly or infeasible to obtain. We show that events in one power system can be accurately detected by reusing a small number of carefully selected labeled PMU data from another without the need for additional labeling. Our transfer learning-based approach outperforms alternative state-of-the-art conventional machine learning (ML) methods on a large PMU historical dataset. We demonstrate this approach with a use case of detecting events from historical PMU data recorded in the Eastern Interconnection in the USA by using similar labeled PMU data from the Western Interconnection. This technique may be propagated to other situations where some of the events’ data from one power system may be applied to enhance learning in another.

The Gaussian Conditional Random Fields (GCRF) algorithm and its extensions are used for machine learning regression problems in which the attributes of objects and the correlation between objects should be considered when making predictions. These algorithms can be applied in different domains where problems can be seen as graphs, but their implementation requires complex calculations and good programming skills. This paper presents an open source software package that includes a tool with graphical user interface (GCRFs tool) and Java library (GCRFs library). GCRFs tool is software that integrates various GCRF-based algorithms and supports training and testing of those algorithms on real-world datasets. The main goal of GCRFs tool is to provide a straightforward and user-friendly graphical user interface that will simplify the use of GCRF-based algorithms. GCRFs Java library contains basic classes for GCRF concepts and can be used by researchers who have experience in Java programming. Also, this paper presents the results of a pilot usability evaluation of the GCRFs tool, where the software was evaluated with expert and non-expert users. This evaluation gave us detailed insight into the experiences and opinions of the users and helped us outline priorities for future development.

Youth anxiety disorders are highly prevalent and associated with considerable concurrent functional impairments. According to the State of the World’s Children report, 13% of youth between 10 and 19 years old have a diagnosed mental health disorder, 40% of which are anxious and depressive disorders. In a typical longitudinal anxiety clinical study, many explanatory variables are observed in a few patients. As patients drop or miss appointments, collected data has a high missing rate in explanatory and predicted variables. We suggest using machine learning methods to improve understanding of treatments and prediction of outcomes in such studies. We propose machine learning-based imputation for understanding youth anxiety data containing features with high missing rates. In the dataset used, the missing rate of features is up to 80%, making them impossible to use in traditional analysis. Our results show that the proposed iterative imputation with a bag of elastic net regressions imputes missing data better than traditional imputation methods and allow for the best prediction result. We investigate imputation and prediction performance change when using jointly data from multiple studies, where each study has a different bias and missing rate. Leveraging joint dataset allows for predicting the therapy outcome in longitudinal studies with few patients. Additionally, we can now impute or predict features and diagnoses not reported by the clinical study. In conducted experiments, pooling data from nine different studies resulted in 9.3% smaller imputation and 33% lower prediction errors, respectively. Results have higher confidence than when studies are considered separately. We also explored the performance of imputation and prediction in the domain adaptation case of withdrawn patients, in which 50% improvement is obtained when data from all studies are used to impute and train the model.

Research on classifying time-series based on subsequences, known as shapelets, has attracted considerable interest in the community. Most existing shapelet-based time-series classification approaches neglect the temporal dependencies among extracted shapelets. Recently, shapelet-orders that encode the temporal dependencies among pairwise shapelets were shown to be informative features. However, based on a random selection of candidate shapelets, the state-of-the-art model does not guarantee optimal shapelets selection. This, in turn, may lead to inferior quality shapelet-orders. Learning shapelets, instead of searching, guarantees near-optimal shapelets thus decreasing generalization error. However, the costly initialization approach for learning generalized shapelets significantly limits its scalability in large time-series datasets. We address the problem of leveraging temporal dependencies among generalized shapelets from randomly initialized subsequences by jointly learning from the shapelet-transform space and the shapelet-order space. The underlying hypothesis is that leveraging the temporal dependency information of generalized shapelets improves the classification performance. Furthermore, introducing a randomized subsequence initialization for learning generalized shapelets allows a more scalable shapelet learning approach. The proposed model was significantly more accurate and faster than the baseline alternatives when evaluated on both synthetic and real-world time-series datasets.

The question of graph similarity or graph distinguishability arises often in natural systems and their analysis over graphical networks. In many domains, graph similarity is used for graph classification, outlier detection or the identification of distinguished interaction patterns. Several methods have been proposed on how to address this topic, but graph comparison still presents many challenges. Recently, information physics has emerged as a promising theoretical foundation for complex networks. In many applications, it has been demonstrated that natural complex systems exhibit features that can be described and interpreted by measures typically applied in quantum mechanical systems. Therefore, a natural starting point for the identification of network similarity measures is information physics and a series of measures of distance for quantum states. In this work, we report experiments on synthetic and real-world data sets, and compare quantum-inspired measures to a series of state-of-the-art and well-established methods of graph distinguishability. We show that quantum-inspired methods satisfy the mathematical and intuitive requirements for graph similarities, while offering high interpretability.

Most of the natural systems encountered in all kinds of disciplines consist of a set of elementary units connected by relationships of different kinds. These complex systems are commonly described in terms of networks, where nodes represent the entities and links represent their interactions. As multiple types of distinct interactions are often observed, these systems are described as multiplex networks where the different types of interactions between the nodes constitute the different layers of the network. The ever-increasing size of these networks introduces new computational challenges and is therefore imperative to be able to eliminate the redundant or irrelevant edges of a network and create a summary that maintains the intrinsic properties of the original network, with respect to the overall structure of the system. In this work, we present a summarization technique for multiplex networks designed to maintain the structural characteristics of such complex systems by utilizing the intrinsic multiplex structure of the network and taking into consideration the inter-connectivity of the various graph layers. We validate our approach on real-world systems from different domains and show that our approach allows for the creation of more compact summaries, with minimum change of the structure evaluation measures, when compared to baseline methods that aggregate contributions of multiple types of interactions.

Digital media companies typically collect rich data in the form of sequences of online user activities. Such data is used in various applications, involving tasks ranging from click or conversion prediction to recommendation or user segmentation. Nonetheless, each application depends upon specialized feature engineering that requires a lot of effort and typically disregards the time-varying nature of the online user behavior. Learning time-preserving vector representations of users (user embeddings), irrespective of a specific task, would save redundant effort and potentially lead to higher embedding quality. To that end, we address the limitations of the current state-of-the-art self-supervised methods for task-independent (unsupervised) sequence embedding, and propose a novel Time-Aware Sequential Autoencoder (TASA) that accounts for the temporal aspects of sequences of activities. The generated embeddings are intended to be readily accessible for many problem formulations and seamlessly applicable to desired tasks, thus sidestepping the burden of task-driven feature engineering. The proposed TASA shows improvements over alternative self-supervised models in terms of sequence reconstruction. Moreover, the embeddings generated by TASA yield increases in predictive performance on both proprietary and public data. It also achieves comparable results to supervised approaches that are trained on individual tasks separately and require substantially more computational effort. TASA has been incorporated within a pipeline designed to provide time-aware user embeddings as a service, and the use of its embeddings exhibited lifts in conversion prediction AUC on four audiences.

OBJECTIVE: We sought to predict if patients with type 2 diabetes mellitus (DM2) would develop 10 selected complications. Accurate prediction of complications could help with more targeted measures that would prevent or slow down their development. MATERIALS AND METHODS: Experiments were conducted on the Healthcare Cost and Utilization Project State Inpatient Databases of California for the period of 2003 to 2011. Recurrent neural network (RNN) long short-term memory (LSTM) and RNN gated recurrent unit (GRU) deep learning methods were designed and compared with random forest and multilayer perceptron traditional models. Prediction accuracy of selected complications were compared on 3 settings corresponding to minimum number of hospitalizations between diabetes diagnosis and the diagnosis of complications. RESULTS: The diagnosis domain was used for experiments. The best results were achieved with RNN GRU model, followed by RNN LSTM model. The prediction accuracy achieved with RNN GRU model was between 73% (myocardial infarction) and 83% (chronic ischemic heart disease), while accuracy of traditional models was between 66% - 76%. DISCUSSION: The number of hospitalizations was an important factor for the prediction accuracy. Experiments with 4 hospitalizations achieved significantly better accuracy than with 2 hospitalizations. To achieve improved accuracy deep learning models required training on at least 1000 patients and accuracy significantly dropped if training datasets contained 500 patients. The prediction accuracy of complications decreases over time period. Considering individual complications, the best accuracy was achieved on depressive disorder and chronic ischemic heart disease. CONCLUSIONS: The RNN GRU model was the best choice for electronic medical record type of data, based on the achieved results.

In this study, a procedure is proposed for surveying public opinion from big social media domain-specific textual data to minimize the difficulties associated with modeling public behavior. Strategies for labeling posts relevant to a topic are discussed. A two-part framework is proposed in which semi-automatic labeling is applied to a small subset of posts, referred to as the "seed" in further text. This seed is used as bases for semi-supervised labeling of the rest of the data. The hypothesis is that the proposed method will achieve better labeling performance than existing classification models when applied to small amounts of labeled data. The seed is labeled using posts of users with a known and consistent view on the topic. A semi-supervised multi-class prediction model labels the remaining data iteratively. In each iteration, it adds context-label pairs to the training set if softmax-based label probabilities are above the threshold. The proposed method is characterized on four datasets by comparison to the three popular text modeling algorithms (n-grams + tfidf, fastText, VDCNN) for different sizes of labeled seeds (5,000 and 50,000 posts) and for several label-prediction significance thresholds. Our proposed semi-supervised method outperformed alternative algorithms by capturing additional contexts from the unlabeled data. The accuracy of the algorithm was increasing by (3-10%) when using a larger fraction of data as the seed. For the smaller seed, lower label probability threshold was clearly a better choice, while for larger seeds no predominant threshold was observed. The proposed framework, using fastText library for efficient text classification and representation learning, achieved the best results for a smaller seed, while VDCNN wrapped in the proposed framework achieved the best results for the bigger seed. The performance was negatively influenced by the number of classes. Finally, the model was applied to characterize a biased dataset of opinions related to gun control/rights advocacy. The proposed semi-automatic seed labeling is used to label 8,448 twitter posts of 171 advocates for guns control/rights. On this application, our approach performed better than existing models and it achieves 96.5% accuracy and 0.68 F1 score.

Temporal graph regression is a frequently encountered research problem in many studies of graph analytics. A temporal graph is a sequence of attributed graphs where node features and target variables change over time, but network structure stays constant. The task of temporal graph regression is to predict the target variables associated with nodes at future time-points given historical snapshots of the graph. Existing methods tackle this problem mostly by conducting structured regression for all target variables. However, those methods have limited performance due to redundant information. Although several techniques have been proposed recently to learn lower dimensional embedding for the target space, the problem of how to effectively exploit the structure of the temporal graph in such embeddings is still unsolved. Other recent works only study node embedding of the stationary graphs only, and this is not applicable to temporal attributed graphs. In this paper, we introduced a Structure-Aware Intrinsic Representation Learning model (SAIRL) to jointly learn lower dimensional embeddings of the target space and feature space via structure-aware graph abstraction and feature-aware target embedding learning. To solve this problem, we have developed a derivative-free block coordinate descent algorithm with closed-form solutions. To characterize the quality of embedding-based learned with SAIRL, we conducted extensive experiments on a variety of different real-world temporal graphs. The results indicate that the proposed method can be more accurate than the state-of-the-art embedding learning methods, regardless of regressors.

Highly discriminative short time-series subsequences, known as shapelets, are used to classify a time-series. The existing shapelet-based methods for time-series classification assume that shapelets are independent of each other. However, they neglect temporal dependencies among pairs of shapelets, which are informative features that exist in many applications. Within this new framework, we explore a scheme to extract informative orders among shapelets by considering the time gap between two shapelets. In addition, we propose a novel model, Pairwise Shapelet-Orders Discovery, which extracts both informative shapelets and shapelet-orders and incorporates the shapelet-transformed space with shapelet-order space for time-series classification. The hypothesis of the study is that the extracted orders could increase the confidence of the prediction and further improves the classification performance. The results of extensive experiments conducted on 75 univariate and 6 multivariate real-world datasets provide evidence that the proposed model could significantly improve accuracy on average over baseline methods.

OBJECTIVE: Clinical trials, prospective research studies on human participants carried out by a distributed team of clinical investigators, play a crucial role in the development of new treatments in health care. This is a complex and expensive process where investigators aim to enroll volunteers with predetermined characteristics, administer treatment(s), and collect safety and efficacy data. Therefore, choosing top-enrolling investigators is essential for efficient clinical trial execution and is 1 of the primary drivers of drug development cost. MATERIALS AND METHODS: To facilitate clinical trials optimization, we propose DeepMatch (DM), a novel approach that builds on top of advances in deep learning. DM is designed to learn from both investigator and trial-related heterogeneous data sources and rank investigators based on their expected enrollment performance on new clinical trials. RESULTS: Large-scale evaluation conducted on 2618 studies provides evidence that the proposed ranking-based framework improves the current state-of-the-art by up to 19% on ranking investigators and up to 10% on detecting top/bottom performers when recruiting investigators for new clinical trials. DISCUSSION: The extensive experimental section suggests that DM can provide substantial improvement over current industry standards in several regards: (1) the enrollment potential of the investigator list, (2) the time it takes to generate the list, and (3) data-informed decisions about new investigators. CONCLUSION: Due to the great significance of the problem at hand, related research efforts are set to shift the paradigm of how investigators are chosen for clinical trials, thereby optimizing and automating them and reducing the cost of new therapies.

Attaining the proper balance between underfitting and overfitting is one of the central challenges in machine learning. It has been approached mostly by deriving bounds on generalization risks of learning algorithms. Such bounds are, however, rarely controllable. In this study, a novel bias-variance balancing objective function is introduced in order to improve generalization performance. By utilizing distance correlation, this objective function is able to indirectly control a stability-based upper bound on a model's expected true risk. In addition, the Generalization-Aware Collaborative Ensemble Regressor (GLACER) is developed, a model that bags a crowd of structured regression models, while allowing them to collaborate in a fashion that minimizes the proposed objective function. The experimental results on both synthetic and real-world data indicate that such an objective enhances the overall model's predictive performance. When compared against a broad range of both traditional and structured regression models GLACER was ~10-56% and ~49-99% more accurate for the task of predicting housing prices and hospital readmissions, respectively.

Optimization of patient throughput and wait time in emergency departments (ED) is an important task for hospital systems. For that reason, Emergency Severity Index (ESI) system for patient triage was introduced to help guide manual estimation of acuity levels, which is used by nurses to rank the patients and organize hospital resources. However, despite improvements that it brought to managing medical resources, such triage system greatly depends on nurse's subjective judgment and is thus prone to human errors. Here, we propose a novel deep model based on the word attention mechanism designed for predicting a number of resources an ED patient would need. Our approach incorporates routinely available continuous and nominal (structured) data with medical text (unstructured) data, including patient's chief complaint, past medical history, medication list, and nurse assessment collected for 338,500 ED visits over three years in a large urban hospital. Using both structured and unstructured data, the proposed approach achieves the AUC of ∼ 88% for the task of identifying resource intensive patients (binary classification), and the accuracy of ∼ 44% for predicting exact category of number of resources (multi-class classification task), giving an estimated lift over nurses' performance by 16% in accuracy. Furthermore, the attention mechanism of the proposed model provides interpretability by assigning attention scores for nurses' notes which is crucial for decision making and implementation of such approaches in the real systems working on human health.

Rapid expansion of mobile devices has brought an unprecedented opportunity for mobile operators and content publishers to reach many users at any point in time. Understanding usage patterns of mobile applications (apps) is an integral task that precedes advertising efforts of providing relevant recommendations to users. However, this task can be very arduous due to the unstructured nature of app data, with sparseness in available information. This study proposes a novel approach to learn representations of mobile user actions using Deep Memory Networks. We validate the proposed approach on millions of app usage sessions built from large scale feeds of mobile app events and mobile purchase receipts. The empirical study demonstrates that the proposed approach performed better compared to several competitive baselines in terms of recommendation precision quality. To the best of our knowledge this is the first study analyzing app usage patterns for purchase recommendation.

Zero-shot learning has tremendous application value in complex computer vision tasks, e.g. image classification, localization, image captioning, etc., for its capability of transferring knowledge from seen data to unseen data. Many recent proposed methods have shown that the formulation of a compatibility function and its generalization are crucial for the success of a zero-shot learning model. In this paper, we formulate a softmax-based compatibility function, and more importantly, propose a regularized empirical risk minimization objective to optimize the function parameter which leads to a better model generalization. In comparison to eight baseline models on four benchmark datasets, our model achieved the highest average ranking. Our model was effective even when the training set size was small and significantly outperforming an alternative state-of-the-art model in generalized zero-shot recognition tasks.

Recovering network connectivity structure from high-dimensional observations is of increasing importance in statistical learning applications. A prominent approach is to learn a Sparse Gaussian Markov Random Field by optimizing regularized maximum likelihood, where the sparsity is induced by imposing L1 norm on the entries of a precision matrix. In this article, we shed light on an alternative objective, where instead of precision, its Cholesky factor is penalized by the L1 norm. We show that such an objective criterion possesses attractive properties that allowed us to develop a very fast Scale-Free Networks Estimation Through Cholesky factorization (SNETCH) optimization algorithm based on coordinate descent, which is highly parallelizable and can exploit an active set approach. The approach is particularly suited for problems with structures that allow sparse Cholesky factor, an important example being scale-free networks. Evaluation on synthetically generated examples and high-impact applications from a biomedical domain of up to more than 900,000 variables provides evidence that for such tasks the SNETCH algorithm can learn the underlying structure more accurately, and an order of magnitude faster than state-of-the-art approaches based on the L1 penalized precision matrix.

Computer simulations based on the finite element method represent powerful tools for modeling blood flow through arteries. However, due to its computational complexity, this approach may be inappropriate when results are needed quickly. In order to reduce computational time, in this paper, we proposed an alternative machine learning based approach for calculation of wall shear stress (WSS) distribution, which may play an important role in mechanisms related to initiation and development of atherosclerosis. In order to capture relationships between geometric parameters, blood density, dynamic viscosity and velocity, and WSS distribution of geometrically parameterized abdominal aortic aneurysm (AAA) and carotid bifurcation models, we proposed multivariate linear regression, multilayer perceptron neural network and Gaussian conditional random fields (GCRF). Results obtained in this paper show that machine learning approaches can successfully predict WSS distribution at different cardiac cycle time points. Even though all proposed methods showed high potential for WSS prediction, GCRF achieved the highest coefficient of determination (0.930-0.948 for AAA model and 0.946-0.954 for carotid bifurcation model) demonstrating benefits of accounting for spatial correlation. The proposed approach can be used as an alternative method for real time calculation of WSS distribution.

To improve the quality of healthcare planning, healthcare systems face challenges in identifying clusters of similar hospitals while considering varying factors. Clustering hospitals based on their admission behavior would be helpful whereas diagnosis of patients is vital in understanding variation in admission. Therefore, grouping hospitals that show similar behavior on their admission distribution while considering similarity among disease symptoms in admission is the objective of our study. This is achieved by a Disease Network of Hospital Networks model which is used to represent hospital admission distribution of multiple diseases as different hospital networks that correspond to disease nodes in a top-layer disease network. This disease network that was extracted from the Human Symptoms Disease Network models the similarity among different disease-specific hospital networks. We assume that disease-specific hospital networks have different underlying clustering structure while share the same underlying clustering structure if corresponding diseases share similar symptoms. Experiments were conducted on more than 14 million electronic health records of monthly admission of 160 diseases over 4 years at 301 hospitals in California. Results of clustering 160 disease-specific hospitals networks that share similar symptoms among corresponding diseases show consistent behavior among these networks when similarity among diseases is considered in clustering process. Patterns of consistent behavior were lacking in results when similarity among diseases is not considered.

Sepsis is a serious, life-threatening condition that presents a growing problem in medicine and health-care. It is characterized by quick progression and high variability in the disease manifestation, so treatment should be personalized and tailored to fit individual characteristics of a particular subject. That requires close monitoring of the patients state and reliable predictions of how the targeted therapy will affect sepsis progression over time. We have characterized predictive capabilities of a graph-based structured regression approach under hemoadsorption therapy by using a computational model of sepsis biomarker progression in rats. Results suggests that an extension of the model representational power by using a dense graph and multiple-step predictors increases predictive accuracy, allowing more appropriate choice of treatment.

Learning the sparse Gaussian Markov Random Field, or conversely, estimating the sparse inverse covariance matrix is an approach to uncover the underlying dependency structure in data. Most of the current methods solve the problem by optimizing the maximum likelihood objective with a Laplace prior L1 on entries of a precision matrix. We propose a novel objective with a regularization term which penalizes an approximate product of the Cholesky decomposed precision matrix. This new reparametrization of the penalty term allows efficient coordinate descent optimization, which in synergy with an active set approach results in a very fast and efficient method for learning the sparse inverse covariance matrix. We evaluated the speed and solution quality of the newly proposed SCHL method on problems consisting of up to 24,840 variables. Our approach was several times faster than three state-of-the-art approaches. We also demonstrate that SCHL can be used to discover interpretable networks, by applying it to a high impact problem from the health informatics domain.

BACKGROUND: Feature selection, aiming to identify a subset of features among a possibly large set of features that are relevant for predicting a response, is an important preprocessing step in machine learning. In gene expression studies this is not a trivial task for several reasons, including potential temporal character of data. However, most feature selection approaches developed for microarray data cannot handle multivariate temporal data without previous data flattening, which results in loss of temporal information. We propose a temporal minimum redundancy - maximum relevance (TMRMR) feature selection approach, which is able to handle multivariate temporal data without previous data flattening. In the proposed approach we compute relevance of a gene by averaging F-statistic values calculated across individual time steps, and we compute redundancy between genes by using a dynamical time warping approach. RESULTS: The proposed method is evaluated on three temporal gene expression datasets from human viral challenge studies. Obtained results show that the proposed method outperforms alternatives widely used in gene expression studies. In particular, the proposed method achieved improvement in accuracy in 34 out of 54 experiments, while the other methods outperformed it in no more than 4 experiments. CONCLUSION: We developed a filter-based feature selection method for temporal gene expression data based on maximum relevance and minimum redundancy criteria. The proposed method incorporates temporal information by combining relevance, which is calculated as an average F-statistic value across different time steps, with redundancy, which is calculated by employing dynamical time warping approach. As evident in our experiments, incorporating the temporal information into the feature selection process leads to selection of more discriminative features.

Increased availability of Electronic Health Record (EHR) data provides unique opportunities for improving the quality of health services. In this study, we couple EHRs with the advanced machine learning tools to predict three important parameters of healthcare quality. More specifically, we describe how to learn low-dimensional vector representations of patient conditions and clinical procedures in an unsupervised manner, and generate feature vectors of hospitalized patients useful for predicting their length of stay, total incurred charges, and mortality rates. In order to learn vector representations, we propose to employ state-of-the-art language models specifically designed for modeling co-occurrence of diseases and applied clinical procedures. The proposed model is trained on a large-scale EHR database comprising more than 35 million hospitalizations in California over a period of nine years. We compared the proposed approach to several alternatives and evaluated their effectiveness by measuring accuracy of regression and classification models used for three predictive tasks considered in this study. Our model outperformed the baseline models on all tasks, indicating a strong potential of the proposed approach for advancing quality of the healthcare system.

Descriptive data analysis is used for mining spatial and temporal patterns from ski lift entrance data. The data, collected through radio-frequency identification scanners, cover one skiing season with approximately 1.2 million recorded ski lift transportations. Cluster analysis was performed on ten subsamples, and the obtained clusters were cross-validated. Several types of skier behavior were found. Temporal clustering revealed that skier patterns differ according to time of maximal performance and length of stay in the ski lift transportation system. Spatial clustering revealed that it is reasonable to have as many clusters as ski lifts in a ski resort, since skiers tend to choose a dominant ski lift during a skier-day. The detected patterns reveal valuable information that can be used for potential improvement of products and services offered by ski resorts.

Gaussian Conditional Random Fields (GCRF) are atype of structured regression model that incorporatesmultiple predictors and multiple graphs. This isachieved by defining quadratic term feature functions inGaussian canonical form which makes the conditionallog-likelihood function convex and hence allows findingthe optimal parameters by learning from data. In thiswork, the parameter space for the GCRF model is extendedto facilitate joint modelling of positive and negativeinfluences. This is achieved by restricting the modelto a single graph and formulating linear bounds on convexitywith respect to the models parameters. In addition,our formulation for the model using one networkallows calculating gradients much faster than alternativeimplementations. Lastly, we extend the model onestep farther and incorporate a bias term into our linkweight. This bias is solved as part of the convex optimization.Benefits of the proposed model in terms ofimproved accuracy and speed are characterized on severalsynthetic graphs with 2 million links as well as on ahospital admissions prediction task represented as a humandisease-symptom similarity network correspondingto more than 35 million hospitalization records inCalifornia over 9 years.

In long-term forecasting it is important to estimate the confidence of predictions, as they are often affected by errors that are accumulated over the prediction horizon. To address this problem, an effective novel iterative method is developed for Gaussian structured learning models in this study for propagating uncertainty in temporal graphs by modeling noisy inputs. The proposed method is applied for three long-term (up to 8 years ahead) structured regression problems on real-world evolving networks from the health and climate domains. The obtained empirical results and use case analysis provide evidence that the new approach allows better uncertainty propagation as compared to published alternatives.

Structured regression has been successfully used in many applications where explanatory and response variables are inter-correlated, such as in weighted attributed networks. One of structured models, Gaussian Conditional Random Fields (GCRF), utilizing multiple unstructured models to learn the non-linear relationships between node attributes and the structured response variable, achieves high prediction accuracy. However, it does not scale well with large networks. We propose a novel model, called Scalable Approximate GCRF (SA-GCRF), which integrates weighted attributed network compression with GCRF, with the aim of making GCRF applicable to large networks. The model consists of three steps: first, it compresses a network into a smaller one by generalizing nodes into supernodes and edges into superedges; then, it applies GCRF to the reduced network; and finally, it unfolds the predicted response variables into the original nodes. Our hypothesis is that the reduced network maintains most information of the original network such that the loss in prediction accuracy obtained by GCRF on the reduced network is minor. The comprehensive experimental results indicate that SA-GCRF was 150-520 times faster than standard GCRF and 11-29 times faster than state-of-the-art UmGCRF on large networks, and provided regression results where GCRF and UmGCRF were not applicable. Furthermore, SA-GCRF achieved a similar regression accuracy, 0.76, to the one obtained from the original real-world weighted attributed citation network, even after compressing the network to 10% of its size.

It is of high interest for a company to identify customers expected to bring the largest profit in the upcoming period. Knowing as much as possible about each customer is crucial for such predictions. However, their demographic data, preferences, and other information that might be useful for building loyalty programs is often missing. Additionally, modeling relations among different customers as a network can be beneficial for predictions at an individual level, as similar customers tend to have similar purchasing patterns. We address this problem by proposing a robust framework for structured regression on deficient data in evolving networks with a supervised representation learning based on neural features embedding. The new method is compared to several unstructured and structured alternatives for predicting customer behavior (e.g. purchasing frequency and customer ticket) on user networks generated from customer databases of two companies from different industries. The obtained results show 4% to 130% improvement in accuracy over alternatives when all customer information is known. Additionally, the robustness of our method is demonstrated when up to 80% of demographic information was missing where it was up to several folds more accurate as compared to alternatives that are either ignoring cases with missing values or learn their feature representation in an unsupervised manner.

Artificial intelligence has achieved superhuman performance in a variety of tasks. Unfortunately, this is often done without interpretable methods. In medicine, it is not sufficient to have an algorithm with maximum accuracy. The methods must be evaluated by experts. Our motivating task is to detect features of the eye using unlabeled data. We detect features using a Bayesian changepoint model. Changepoint detection can perform object recognition when applied to two dimensional feature spaces such as images. We present a formula for detecting any shape with the changepoint model. We then extend it to multiple features in order to capture the color of an object. The work presented here has the ability to incorporate prior information, the ability to handle images of varying granularity, can provide confidence estimates on object features in an image. It will serve as the foundation for a classification method with interpretable results.

In many applications, including climate science, power systems, and remote sensing, multiple input variables are observed for each output variable and the output variables are dependent. Several methods have been proposed to improve prediction by learning the conditional distribution of the output variables. However, when the relationship between the raw features and the outputs is nonlinear, the existing methods cannot capture both the nonlinearity and the underlying structure well. In this study, we propose a structured model containing hidden variables, which are nonlinear functions of inputs and which are linearly related with the output variables. The parameters modeling the relationships between the input and hidden variables, between the hidden and output variables, as well as among the output variables are learned simultaneously. To demonstrate the effectiveness of our proposed method, we conducted extensive experiments on eight synthetic datasets and three real-world challenging datasets: forecasting wind power, forecasting solar energy, and forecasting precipitation over U.S. The proposed method was more accurate than state-of-the-art structured regression methods.

Until recently, research in social network analysis was focused on single layer networks, where only one type of links among nodes is considered. This approach does not consider the variety of interactions that exist among nodes, resulting in the loss of a large amount of information. In the last few years there is an advanced interest in multilayer networks analysis, where multiple types of nodal connections are considered jointly. In most approaches however the contributions of the various interactions are averaged, resulting again in the loss of information. In this work we present a structured regression model for node attribute prediction in multilayer networks. Our Gaussian Conditional Random Fields model is designed to maximize the information gained from the use of data with multiple layers of graphical structure. Our model accommodates graphs with layers that share the same set of nodes allowing for missing nodes and unobserved connections. At the same time it models the evolution of such networks over time without requiring the addition of a new layer. We present evidence that this model outperforms the traditionally used one and that it offers predictive accuracy that increases as the number of layers used grows, on both synthetic data and challenging real world applications such as predicting citation count and sepsis hospitalization admission rate at all hospitals in California.

BACKGROUND: Machine learning models have been adapted in biomedical research and practice for knowledge discovery and decision support. While mainstream biomedical informatics research focuses on developing more accurate models, the importance of data preprocessing draws less attention. We propose the Generalized Logistic (GL) algorithm that scales data uniformly to an appropriate interval by learning a generalized logistic function to fit the empirical cumulative distribution function of the data. The GL algorithm is simple yet effective; it is intrinsically robust to outliers, so it is particularly suitable for diagnostic/classification models in clinical/medical applications where the number of samples is usually small; it scales the data in a nonlinear fashion, which leads to potential improvement in accuracy. RESULTS: To evaluate the effectiveness of the proposed algorithm, we conducted experiments on 16 binary classification tasks with different variable types and cover a wide range of applications. The resultant performance in terms of area under the receiver operation characteristic curve (AUROC) and percentage of correct classification showed that models learned using data scaled by the GL algorithm outperform the ones using data scaled by the Min-max and the Z-score algorithm, which are the most commonly used data scaling algorithms. CONCLUSION: The proposed GL algorithm is simple and effective. It is robust to outliers, so no additional denoising or outlier detection step is needed in data preprocessing. Empirical results also show models learned from data scaled by the GL algorithm have higher accuracy compared to the commonly used data scaling algorithms.

BACKGROUND: Existing feature selection methods typically do not consider prior knowledge in the form of structural relationships among features. In this study, the features are structured based on prior knowledge into groups. The problem addressed in this article is how to select one representative feature from each group such that the selected features are jointly discriminating the classes. The problem is formulated as a binary constrained optimization and the combinatorial optimization is relaxed as a convex-concave problem, which is then transformed into a sequence of convex optimization problems so that the problem can be solved by any standard optimization algorithm. Moreover, a block coordinate gradient descent optimization algorithm is proposed for high dimensional feature selection, which in our experiments was four times faster than using a standard optimization algorithm. RESULTS: In order to test the effectiveness of the proposed formulation, we used microarray analysis as a case study, where genes with similar expressions or similar molecular functions were grouped together. In particular, the proposed block coordinate gradient descent feature selection method is evaluated on five benchmark microarray gene expression datasets and evidence is provided that the proposed method gives more accurate results than the state-of-the-art gene selection methods. Out of 25 experiments, the proposed method achieved the highest average AUC in 13 experiments while the other methods achieved higher average AUC in no more than 6 experiments. CONCLUSION: A method is developed to select a feature from each group. When the features are grouped based on similarity in gene expression, we showed that the proposed algorithm is more accurate than state-of-the-art gene selection methods that are particularly developed to select highly discriminative and less redundant genes. In addition, the proposed method can exploit any grouping structure among features, while alternative methods are restricted to using similarity based grouping.

The availability of big data [James et al. 2011; Steve 2012] and the emergence of wearable computing [Thad 1996; Alex 2000], network science [Barabasi 2002], and computational social science [Hanna 2016; Watts and Strogatz 1998] as areas of inquiry has been revolutionizing the landscape of how we decipher our lives, our social interactions, and our day-to-day activities.This well-connected world has promised novel requirements on transforming healthcare from reactive and hospital-centered, to preventive, proactive, evidence-based, person-centered, and focused on well-being rather than ailment recovery.A multitude of various types of data are involved in this broad context of healthcare, including the following:-Clinical data [Prather et al. 1997; Riccardo and Zupan 2008], mainly the patient records from clinical institutions, such as medical imaging, patient electronic health records, clinical trial data, etc. -Genotype data [Eibe et al. 2004; Leslie et al. 1999], basically the genetic makeups of the individuals, such as DNA, protein, etc. -Social media data [Reza et al. 2014; Sitaram and Huberman 2010], which is the information the individuals posted on online social platforms such as Facebook, Twitter, PatientsLikeMe, etc. -Environmental sensory data [Ruchi and Bhatia 2010], which is the information sampled from the surrounding environment where the individuals are living in, such as air pollution and humidity information.-Behavioral and sentiment data [Bo and Lee 2008], which could be the data recorded by the wearable devices on patient's activities.-Mobile data [Miller and Han 2009; Fosca and Pedreschi 2008], which is sampled from individuals' mobile devices.Integrating multiple types of information to make people healthier is also a problem of vital importance that requires collective effort from different parties, where data mining plays a pivotal role.Toward this aim, the National Science Foundation of United

Gene expression data are widely used in classification tasks for medical diagnosis. Data scaling is recommended and helpful for learning the classification models. In this study, we propose a data scaling algorithm to transform the data uniformly to an appropriate interval by learning a generalized logistic function to fit the empirical cumulative density function of the data. The proposed algorithm is robust to outliers, and experimental results show that models learned using data scaled by the proposed algorithm generally outperform the ones using min-max mapping and z-score which are currently the most commonly used data scaling algorithms.

High false alarm rates in intensive care units (ICUs) cause desensitization among care providers, thus risking patients' lives. Providing early detection of true and false cardiac arrhythmia alarms can alert hospital personnel and avoid alarm fatigue, so that they can act only on true life-threatening alarms, hence improving efficiency in ICUs. However, suppressing false alarms cannot be an excuse to suppress true alarm detection rates. In this study, we investigate a cost-sensitive approach for false alarm suppression while keeping near perfect true alarm detection rates. Our experiments on two life threatening cardiac arrhythmia datasets from Physionet's MIMIC II repository provide evidence that the proposed method is capable of identifying patterns that can distinguish false and true alarms using on average 60% of the available time series' length. Using temporal uncertainty estimates of time series predictions, we were able to estimate the confidence in our early classification predictions, therefore providing a cost-sensitive prediction model for ECG signal classification. The results from the proposed method are interpretable, providing medical personnel a visual verification of the predicted results. In conducted experiments, moderate false alarm suppression rates were achieved (34.29% for Asystole and 20.32% for Ventricular Tachycardia) while keeping near 100% true alarm detection, outperforming the state-of-the-art methods, which compromise true alarm detection rate for higher false alarm suppression rate, on these challenging applications.

Conditional probabilistic graphical models provide a powerful framework for structured regression in spatio-temporal datasets with complex correlation patterns. However, in real-life applications a large fraction of observations is often missing, which can severely limit the representational power of these models. In this paper we propose a Marginalized Gaussian Conditional Random Fields (m-GCRF) structured regression model for dealing with missing labels in partially observed temporal attributed graphs. This method is aimed at learning with both labeled and unlabeled parts and effectively predicting future values in a graph. The method is even capable of learning from nodes for which the response variable is never observed in history, which poses problems for many state-of-the-art models that can handle missing data. The proposed model is characterized for various missingness mechanisms on 500 synthetic graphs. The benefits of the new method are also demonstrated on a challenging application for predicting precipitation based on partial observations of climate variables in a temporal graph that spans the entire continental US. We also show that the method can be useful for optimizing the costs of data collection in climate applications via active reduction of the number of weather stations to consider. In experiments on these real-world and synthetic datasets we show that the proposed model is consistently more accurate than alternative semi-supervised structured models, as well as models that either use imputation to deal with missing values or simply ignore them altogether.

We present a Gaussian conditional random field model for the aggregation of aerosol optical depth (AOD) retrievals from multiple satellite instruments into a joint retrieval. The model provides aggregated retrievals with higher accuracy and coverage than any of the individual instruments while also providing an estimation of retrieval uncertainty. The proposed model finds an optimal temporally smoothed combination of individual retrievals that minimizes the root-mean-squared error of AOD retrieval. We evaluated the model on five years (2006-2010) of satellite data over North America from five instruments (Aqua and Terra MODIS, MISR, SeaWiFS, and the Ozone Monitoring Instrument), collocated with ground-based Aerosol Robotic Network ground-truth AOD readings, clearly showing that the aggregation of different sources leads to improvements in the accuracy and coverage of AOD retrievals.

Predictive insights on extreme and rare events are important across multiple disciplines ranging from hydrology, climate, and remote sensing to finance and security. Characterizing the dependence of extremes on covariates can help in identification of plausible causal drivers and may even inform predictive modeling. However, despite progress in the incorporation of covariates in the statistical theory of extremes and in sparse covariate discovery algorithms, progress has been limited for high-dimensional data where the number of covariates is large. In this paper, we propose a general-purpose sparse Bayesian framework for covariate discovery based on a Poisson description of extremes frequency and a hierarchical Bayesian description of a sparse regression model. We obtain posteriors over regression coefficients, which indicate dependence of extremes on the corresponding covariates, using a variational Bayes approximation. Experiments with synthetic data demonstrate the ability of the approach to accurately characterize dependence structures. The method is applied to discover the covariates affecting the frequency of precipitation extremes obtained from station-level observations over nine climatologically homogeneous regions within the continental U.S. The candidate covariates at multiple spatial scales represent station-level as well as regional and seasonal atmospheric condition, indices that attempt to capture large-scale ocean-based climate oscillators and hence natural climate variability, as well as global warming. Our results confirm the dependence structures that may be expected from known precipitation physics and generate novel insights, which can inform physical understanding and perhaps even predictive modeling.

Accurate estimation of what a day in a hospital costs and what the hospital charges is of high interest to many parties, including health care providers, medical insurance companies, health researchers, and uninsured patients. The problem is complex, as the cost-to-charge ratio varies greatly from hospital to hospital and over time. In addition, the cost-to-charge ratio is often not reported, and in such cases group average values from similar hospitals are used. In this study we address the problem of estimating the cost-to-charge ratio at the hospital level by utilizing structured regression on a temporal graph of more than 4,000 hospitals, observed over 8 years, constructed from the National Inpatient Sample database. In the proposed approach, the cost-to-charge estimates at individual hospitals for a certain month obtained by an artificial neural network were used as unstructured components in the Gaussian Conditional Random Fields (GCRF) model. The diagnosis codes of treatments in each hospital were used to create a similarity metric that represents correlation among hospital specializations. The estimates of cost-to-charge ratio obtained using convex optimization of the GCRF parameters on the constructed graph were much better than those relying on group average based cost-to-charge estimates. In addition, cost-to-charge ratio estimates by our GCRF model outperformed regression by nonlinear artificial neural networks.

Early classification of time series is prevalent in many time-sensitive applications such as, but not limited to, early warning of disease outcome and early warning of crisis in stock market. \textcolor{black}{ For example,} early diagnosis allows physicians to design appropriate therapeutic strategies at early stages of diseases. However, practical adaptation of early classification of time series requires an easy to understand explanation (interpretability) and a measure of confidence of the prediction results (uncertainty estimates). These two aspects were not jointly addressed in previous time series early classification studies, such that a difficult choice of selecting one of these aspects is required. In this study, we propose a simple and yet effective method to provide uncertainty estimates for an interpretable early classification method. The question we address here is "how to provide estimates of uncertainty in regard to interpretable early prediction." In our extensive evaluation on twenty time series datasets we showed that the proposed method has several advantages over the state-of-the-art method that provides reliability estimates in early classification. Namely, the proposed method is more effective than the state-of-the-art method, is simple to implement, and provides interpretable results.

When used for structured regression, powerful Conditional Random Fields (CRFs) are typically restricted to modeling effects of interactions among examples in local neighborhoods. Using more expressive representation would result in dense graphs, making these methods impractical for large-scale applications. To address this issue, we propose an effective CRF model with linear scale-up properties regarding approximate learning and inference for structured regression on large, fully connected graphs. The proposed method is validated on real-world large-scale problems of image de-noising and remote sensing. In conducted experiments, we demonstrated that dense connectivity provides an improvement in prediction accuracy. Inference time of less than ten seconds on graphs with millions of nodes and trillions of edges makes the proposed model an attractive tool for large-scale, structured regression problems.

When faced with the task of forming predictions for nodes in a social network, it can be quite difficult to decide which of the available connections among nodes should be used for the best results. This problem is further exacerbated when temporal information is available, prompting the question of whether this information should be aggregated or not, and if not, which portions of it should be used. With this challenge in mind, we propose a novel utilization of variograms for selecting potentially useful relationship types, whose merits are then evaluated using a Gaussian Conditional Random Field model for node attribute prediction of temporal social networks with a multigraph structure. Our flexible model allows for measuring many kinds of relationships between nodes in the network that evolve over time, as well as using those relationships to augment the outputs of various unstructured predictors to further improve performance. The experimental results exhibit the effectiveness of using particular relationships to boost performance of unstructured predictors, show that using other relationships could actually impede performance, and also indicate that while variograms alone are not necessarily sufficient to identify a useful relationship, they greatly help in removing obviously useless measures, and can be combined with intuition to identify the optimal relationships.

Leveraging temporal observations to predict a patient's health state at a future period is a very challenging task. Providing such a prediction early and accurately allows for designing a more successful treatment that starts before a disease completely develops. Information for this kind of early diagnosis could be extracted by use of temporal data mining methods for handling complex multivariate time series. However, physicians usually prefer to use interpretable models that can be easily explained, rather than relying on more complex black-box approaches. In this study, a temporal data mining method is proposed for extracting interpretable patterns from multivariate time series data, which can be used to assist in providing interpretable early diagnosis. The problem is formulated as an optimization based binary classification task addressed in three steps. First, the time series data is transformed into a binary matrix representation suitable for application of classification methods. Second, a novel convex-concave optimization problem is defined to extract multivariate patterns from the constructed binary matrix. Then, a mixed integer discrete optimization formulation is provided to reduce the dimensionality and extract interpretable multivariate patterns. Finally, those interpretable multivariate patterns are used for early classification in challenging clinical applications. In the conducted experiments on two human viral infection datasets and a larger myocardial infarction dataset, the proposed method was more accurate and provided classifications earlier than three alternative state-of-the-art methods.

BACKGROUND: In many biomedical applications, there is a need for developing classification models based on noisy annotations. Recently, various methods addressed this scenario by relaying on unreliable annotations obtained from multiple sources. RESULTS: We proposed a probabilistic classification algorithm based on labels obtained by multiple noisy annotators. The new algorithm is capable of eliminating annotations provided by novice labellers and of providing a more accurate estimate of the ground truth by consensus labelling according to higher quality annotations. The approach is evaluated on text classification and prediction of protein disorder. Our study suggests that the higher levels of accuracy, effectiveness and performance can be achieved by the new method as compared to alternatives. CONCLUSIONS: The proposed method is applicable for meta-learning from multiple existing classification models and noisy annotations obtained by humans. It is particularly beneficial when many annotations are obtained by novice labellers. In addition, the proposed method can provide further characterization of each annotator that can help in developing more accurate classifiers by identifying the most competent annotators for each data instance.

This study considers the problem of feature selection in incomplete data. The intuitive approach is to first impute the missing values, and then apply a standard feature selection method to select relevant features. In this study, we show how to perform feature selection directly, without imputing missing values. We define the objective function of the uncertainty margin-based feature selection method to maximize each instance’s uncertainty margin in its own relevant subspace. In optimization, we take into account the uncertainty of each instance due to the missing values. The experimental results on synthetic and 6 benchmark data sets with few missing values (less than 25%) provide evidence that our method can select the same accurate features as the alternative methods which apply an imputation method first. However, when there is a large fraction of missing values (more than 25%) in data, our feature selection method outperforms the alternatives, which impute missing values first.

Early classification of time series has been receiving a lot of attention as of late, particularly in the context of gene expression. In the biomédical realm, early classification can be of tremendous help, by identifying the onset of a disease before it has time to fully take hold, or determining that a treatment has done its job and can be discontinued. In this paper we present a state-of-the-art model, which we call the Early Classification Model (ECM), that allows for early, accurate, and patient-specific classification of multivariate time series. The model is comprised of an integration of the widely-used HMM and SVM models, which, while not a new technique per se, has not been used for early classification of multivariate time series classification until now. It attained very promising results on the datasets we tested it on: in our experiments based on a published dataset of response to drug therapy in Multiple Sclerosis patients, ECM used only an average of 40% of a time series and was able to outperform some of the baseline models, which needed the full time series for classification.

In order to more accurately predict an individual's health status, in clinical applications it is often important to perform analysis of high-dimensional gene expression data that varies with time. A major challenge in predicting from such temporal microarray data is that the number of biomarkers used as features is typically much larger than the number of labeled subjects. One way to address this challenge is to perform feature selection as a preprocessing step and then apply a classification method on selected features. However, traditional feature selection methods cannot handle multivariate temporal data without applying techniques that flatten temporal data into a single matrix in advance. In this study, a feature selection filter that can directly select informative features from temporal gene expression data is proposed. In our approach we measure the distance between multivariate temporal data from two subjects. Based on this distance, we define the objective function of temporal margin based feature selection to maximize each subject's temporal margin in its own relevant subspace. The experimental results on two real flu data sets provide evidence that our method outperforms the alternatives, which flatten the temporal data in advance.

Acute inflammation is a medical condition which occurs over seconds, minutes or hours and is characterized as a systemic inflammatory response to an infection. Delaying treatment by only one hour decreases patient chance of survival by about 7%. Therefore, there is a critical need for tools that can aid therapy optimization for this potentially fatal condition. Towards this objective we developed a data driven approach for therapy optimization where a predictive model for patients' behavior is learned directly from historical data. As such, the predictive model is incorporated into a model predictive control optimization algorithm to find optimal therapy, which will lead the patient to a healthy state. To save on the cost of clinical trials and potential failure, we evaluated our model on a population of virtual patients capable of emulating the inflammatory response. Patients are treated with two drugs for which dosage and timing are critical for the outcome of the treatment. Our results show significant improvement in percentage of healthy outcomes comparing to previously proposed methods for acute inflammation treatment found in literature and in clinical practice. In particular, application of our method rescued 88% of patients that would otherwise die within 168 hours due to septic or aseptic state. In contrast, the best method from literature rescued only 73% of patients.

Learning from noisy labels obtained from multiple annotators and without access to any true labels is an increasingly important problem in bioinformatics and biomedicine. In our method, this challenge is addressed by iteratively filtering low-quality annotators and estimating the consensus labels based only on the remaining experts that provide higher-quality annotations. Experiments on biomedical text classification and CASP9 protein disorder prediction tasks provide evidence that the proposed algorithm is more accurate than the majority voting and previously developed multi-annotator approaches. The benefit of using the new method is particularly large when low-quality annotators dominate. Moreover, the new algorithm also suggests the most relevant annotators for each instance, thus paving the way for understanding the behaviors of each annotator and building more reliable predictive models for bioinformatics applications.

Genome-wide analysis of single nucleotide polymorphisms (SNP) can potentially be helpful in exploring the role of genetic variability in drug therapy. However, two major problems with such an analysis are the need for a large number of interrogated genomes, and the resulting high-dimensional data where the number of SNPs used as features is much larger than the number of subjects. The aim of this study is to identify informative SNPs associated with clinical efficacy and side effects of domperidone treatment for gastroparesis from DNA microarray experiments by applying our feature selection method, which approximates the Markov Blanket in a kernel-induced space. DNA samples extracted from the saliva of 46 patients treated with domperidone were analyzed using Affymetrix 6.0 SNP microarrays. Experimental evaluations on this SNP microarray dataset provide evidence that our feature selection method can remove useless SNP features more accurately than existing Markov Blanket based alternatives.

We propose a method for designing a partitioning clustering algorithm from reusable components that is suitable for finding the appropriate number of clusters (K) in microarray data. The proposed method is evaluated on 10 datasets (4 syntetic and 6 real-word microarrays) by considering 1008 reusable-component-based algorithms and four normalization methods. The best performing algorithm were reported on every dataset and also rules were identified for designing microarray-specific clustering algorithms. The obtained results indicate that in the majority of cases a data-tailored clustering algorithm design outperforms the results reported in the literature. In addition, data normalization can have an important influence on algorithm performance. The method proposed in this paper gives insights for design of divisive clustering algorithms that can reveal the optimal K in a microarray dataset.

Identifying informative biomarkers from a large pool of candidates is the key step for accurate prediction of an individual's health status. In clinical applications traditional static feature selection methods that flatten the temporal data cannot be directly applied since the patient's observed clinical condition is a temporal multivariate time series where different variables can capture various stages of temporal change in the patient's health status. In this study, in order to identify informative genes in temporal microarray data, a margin based feature selection filter is proposed. The proposed method is based on well-established machine learning techniques without any assumptions about the data distribution. The objective function of temporal margin-based feature selection is defined to maximize each subject's temporal margin in its own relevant subspace. In the objective function, the uncertainty in calculating nearest neighbors is taken into account by considering the change in feature weights in each iteration. A fixed-point gradient descent method is proposed to solve the formulated objective function. The experimental results on both synthetic and real data provide evidence that the proposed method can identify more informative features than the alternatives that flatten the temporal data in advance.

Building prediction models for suggestive knowledge from multiple sources dynamically is of great interest from a clinical decision support point of view. This is valuable in situations where the local clinical data repository does not have sufficient number of records to draw conclusions from. However, due to privacy concerns, hospitals are reluctant to divulge patient records. Consequently, a distributed model building mechanism that can use just the statistics from multiple hospitals' databases is valuable. Our DIDT algorithm builds a model in that fashion. In this study, using National Inpatient Sample (NIS) data for 2009, we demonstrate that DIDT algorithm can be used to help collaboratively build a better decision-making model in situations where hospitals have small number of records that are insufficient to make good local models. Based on 262 attributes used for model building, we showed that 9 collaborating hospitals each with less than 100 cases of hospitalizations related to diabetes were able to achieve 9.9% improvement in accuracies of hospitalization prediction collectively using a distributed model as compared to relying on local models developed on their own. When relying on local risk prediction models for diabetes at these 9 hospitals, 159 of 357 patients were misclassified and prediction was impossible for another 16 patients. Our integrated model reduced the misclassification to 138 effectively providing accurate early diagnostics to 37 additional patients. We also introduce the concept of banding to improve DIDT algorithm so as to logically combine multiple hospitals when large number of hospitals is involved for reduction in cross-validation folds.

We propose a novel semi-supervised learning algorithm, called IMCC, designed for co-training classifiers on single-view datasets. Our method runs the co-training algorithm for a predefined number of times, each time using a different random split of features. Thus, a set of diverse co-training classifiers is created. Each of these classifiers then labels each of the examples for which we want to determine the class label. In this way, each example for classification is assigned multiple labels. We then treat this as a problem of learning from inconsistent and unreliable annotators in a multi-annotator problem setting and estimate the single hidden true label for each example. In experimental results obtained on 25 benchmark datasets of various properties IMCC outperformed five considered alternative methods for co-training on single-view datasets, and resulted in a statistical tie with a Naive Bayes classifier trained using a much larger set of labeled examples.

Discovery of dependence structure between precipitation extremes and other climate variables (covariates) within a smaller spatial and temporal neighborhood is an important step in better understanding the drivers of this complex phenomenon as well as short-term prediction of extremes occurrence. Apart from the inherent spatio-temporal variability of the dependence, it is further complicated by the availability of the covariates at different vertical levels. The above problem can be split into three different sub-problems. Firstly, a spatio-temporal neighborhood of influence has to be discovered, which can be different for different locations. Secondly, the dependence structure between the precipitation extremes and the covariates has to be discovered within this neighborhood and thirdly, it has to be investigated whether this dependence structure can be exploited for any predictive power. Climate scientists have already discovered some physics-based relations between some of the covariates (e.g. temperature, relative humidity, precipitable water etc.) and precipitation extremes. We are exploring data-dependent alternatives for these problems and any possibility of incorporating the physics-based relations into the resulting data model. In particular, we used elastic net-based sparse optimization technique which solves all three problems of neighborhood discovery, covariate dependence discovery and predictive modeling and at the same time maintains the interpretability of the resulting model. Preliminary results look promising and show potential for some interesting knowledge discovery. We are currently exploring non-linear correlations and the alternatives to combine the physics-based relationships into the data model.

BACKGROUND: The relationships between the gene functional similarity and gene expression profile, and between gene function annotation and gene sequence have been studied extensively. However, not much work has considered the connection between gene functions and location of a gene's expression in the mammalian tissues. On the other hand, although unsupervised learning methods have been commonly used in functional genomics, supervised learning cannot be directly applied to a set of normal genes without having a target (class) attribute. RESULTS: Here, we propose a supervised learning methodology to predict pair-wise gene functional similarity from multiplex gene expression maps that provide information about the location of gene expression. The features are extracted from expression maps and the labels denote the functional similarities of pairs of genes. We make use of wavelet features, original expression values, difference and average values of neighboring voxels and other features to perform boosting analysis. The experimental results show that with increasing similarities of gene expression maps, the functional similarities are increased too. The model predicts the functional similarities between genes to a certain degree. The weights of the features in the model indicate the features that are more significant for this prediction. CONCLUSIONS: By considering pairs of genes, we propose a supervised learning methodology to predict pair-wise gene functional similarity from multiplex gene expression maps. We also explore the relationship between similarities of gene maps and gene functions. By using AdaBoost coupled with our proposed weak classifier we analyze a large-scale gene expression dataset and predict gene functional similarities. We also detect the most significant single voxels and pairs of neighboring voxels and visualize them in the expression map image of a mouse brain. This work is very important for predicting functions of unknown genes. It also has broader applicability since the methodology can be applied to analyze any large-scale dataset without a target attribute and is not restricted to gene expressions.

BACKGROUND: Early classification of time series is beneficial for biomedical informatics problems such including, but not limited to, disease change detection. Early classification can be of tremendous help by identifying the onset of a disease before it has time to fully take hold. In addition, extracting patterns from the original time series helps domain experts to gain insights into the classification results. This problem has been studied recently using time series segments called shapelets. In this paper, we present a method, which we call Multivariate Shapelets Detection (MSD), that allows for early and patient-specific classification of multivariate time series. The method extracts time series patterns, called multivariate shapelets, from all dimensions of the time series that distinctly manifest the target class locally. The time series were classified by searching for the earliest closest patterns. RESULTS: The proposed early classification method for multivariate time series has been evaluated on eight gene expression datasets from viral infection and drug response studies in humans. In our experiments, the MSD method outperformed the baseline methods, achieving highly accurate classification by using as little as 40%-64% of the time series. The obtained results provide evidence that using conventional classification methods on short time series is not as accurate as using the proposed methods specialized for early classification. CONCLUSION: For the early classification task, we proposed a method called Multivariate Shapelets Detection (MSD), which extracts patterns from all dimensions of the time series. We showed that the MSD method can classify the time series early by using as little as 40%-64% of the time series' length.

Neural networks have the ability to represent and learn complex regression functions and are very suitable for retrieval of geophysical parameters from remotely sensed data. Neural networks trained to minimize the mean square error are able to estimate the conditional expectation of target variables. In many remote sensing applications, it is also critical to provide estimates of prediction uncertainty. In this paper, we evaluate an approach that, in addition to training a neural network for retrievals, also trains a neural-network-based estimator of retrieval uncertainty. The uncertainty estimator is built under the assumption that uncertainty is a function of input variables. The methodology was evaluated on aerosol-optical-depth retrieval. The data set consists of 38 238 collocated Moderate Resolution Imaging Spectrometer (MODIS) satellite instrument and Aerosol Robotic Network ground-based instrument measurements collected over the entire Earth during two years (in 2005-2006). The results indicate that a neural network ensemble is more accurate than the operational MODIS retrieval algorithm called Collection 5 and that the retrieval uncertainty of the ensemble can be estimated with satisfactory accuracy.

Intrinsically disordered proteins perform a variety of crucial biological functions despite lacking stable tertiary structure under physiological conditions in vitro. State-of-the-art sequence-based predictors of intrinsic disorder are achieving per-residue accuracies over 80%. In a genome-wide study we observed big difference in predicted disorder content between confirmed and putative human proteins, and suspected that this is due to large errors introduced by gene-finding algorithms for putative sequence annotation. To test this hypothesis we trained a predictor to discriminate sequences of real proteins from synthetic sequences that mimic errors of gene finding algorithms. Its application to putative human protein sequences shows that they contain a substantial fraction of incorrectly assigned regions. These regions are predicted to have higher levels of disorder content than correctly assigned regions. Our finding provides first evidence that current practice of predicting disorder content in putative sequences should be reconsidered, as such estimates are biased.

Several external indices that use information not present in the dataset were shown to be useful for evaluation of representative based clustering algorithms. However, such supervised measures are not directly useful for construction of better clustering algorithms when class labels are not provided. We propose a method for identifying internal cluster evaluation measures that use only information present in the dataset and are related to given external indices. We utilize these internal measures for the construction of representative based clustering algorithms. Both identification and utilization steps of the proposed method are enabled by use of a component-based clustering algorithm design. Experiments on 432 algorithms using gene expression data sets provide evidence that some internal measures could be used as surrogates for external indices proposed in the literature. Moreover, the obtained results suggest that internal measures correlated to selected external indices can guide the algorithms toward significantly better cluster models.

Prediction models for multivariate spatio-temporal functions in geosciences are typically developed using supervised learning from attributes collected by remote sensing instruments collocated with the outcome variable provided at sparsely located sites. In such collocated data there are often large temporal gaps due to missing attribute values at sites where outcome labels are available. Our objective is to develop more accurate spatio-temporal predictors by using enlarged collocated data obtained by imputing missing attributes at time and locations where outcome labels are available. The proposed method for large gaps estimation in space and time (called LarGEST) exploits temporal correlation of attributes, correlations among multiple attributes collected at the same time and space, and spatial correlations among attributes from multiple sites. LarGEST outperformed alternative methods in imputing up to 80 % of randomly missing observations at a synthetic spatio-temporal signal and at a model of fluoride content in a water distribution system. LarGEST was also applied for imputing 80 % of nonrandom missing values in data from one of the most challenging Earth science problems related to aerosol properties. Using such enlarged data a predictor of aerosol optical depth is developed that was much more accurate than predictors based on alternative imputation methods when tested rigorously over entire continental US in year 2005. 1

BACKGROUND: A significant number of proteins have been shown to be intrinsically disordered, meaning that they lack a fixed 3 D structure or contain regions that do not posses a well defined 3 D structure. It has also been proven that a protein's disorder content is related to its function. We have performed an exhaustive analysis and comparison of the disorder content of proteins from prokaryotic organisms (i.e., superkingdoms Archaea and Bacteria) with respect to functional categories they belong to, i.e., Clusters of Orthologous Groups of proteins (COGs) and groups of COGs-Cellular processes (Cp), Information storage and processing (Isp), Metabolism (Me) and Poorly characterized (Pc). We also analyzed the disorder content of proteins with respect to various genomic, metabolic and ecological characteristics of the organism they belong to. We used correlations and association rule mining in order to identify the most confident associations between specific modalities of the characteristics considered and disorder content. RESULTS: Bacteria are shown to have a somewhat higher level of protein disorder than archaea, except for proteins in the Me functional group. It is demonstrated that the Isp and Cp functional groups in particular (L-repair function and N-cell motility and secretion COGs of proteins in specific) possess the highest disorder content, while Me proteins, in general, posses the lowest. Disorder fractions have been confirmed to have the lowest level for the so-called order-promoting amino acids and the highest level for the so-called disorder promoters. For each pair of organism characteristics, specific modalities are identified with the maximum disorder proteins in the corresponding organisms, e.g., high genome size-high GC content organisms, facultative anaerobic-low GC content organisms, aerobic-high genome size organisms, etc. Maximum disorder in archaea is observed for high GC content-low genome size organisms, high GC content-facultative anaerobic or aquatic or mesophilic organisms, etc. Maximum disorder in bacteria is observed for high GC content-high genome size organisms, high genome size-aerobic organisms, etc. Some of the most reliable association rules mined establish relationships between high GC content and high protein disorder, medium GC content and both medium and low protein disorder, anaerobic organisms and medium protein disorder, Gammaproteobacteria and low protein disorder, etc. A web site Prokaryote Disorder Database has been designed and implemented at the address http://bioinfo.matf.bg.ac.rs/disorder, which contains complete results of the analysis of protein disorder performed for 296 prokaryotic completely sequenced genomes. CONCLUSIONS: Exhaustive disorder analysis has been performed by functional classes of proteins, for a larger dataset of prokaryotic organisms than previously done. Results obtained are well correlated to those previously published, with some extension in the range of disorder level and clear distinction between functional classes of proteins. Wide correlation and association analysis between protein disorder and genomic and ecological characteristics has been performed for the first time. The results obtained give insight into multi-relationships among the characteristics and protein disorder. Such analysis provides for better understanding of the evolutionary process and may be useful for taxon determination. The main drawback of the approach is the fact that the disorder considered has been predicted and not experimentally established.

Typical data mining algorithms follow a so called “black-box” paradigm, where the logic is hidden from the user not to overburden him. We show that “white-box” algorithms constructed with reusable components design can have significant benefits for r

Studies of intrinsically disordered proteins that lack a stable tertiary structure but still have important biological functions critically rely on computational methods that predict this property based on sequence information. Although a number of fairly successful models for prediction of protein disorder were developed over the last decade, the quality of their predictions is limited by available cases of confirmed disorders. To more reliably estimate protein disorder from protein sequences, an iterative algorithm is proposed that integrates predictions of multiple disorder models without relying on any protein sequences with confirmed disorder annotation. The iterative method alternately provides the maximum a posterior (MAP) estimation of disorder prediction and the maximum-likelihood (ML) estimation of quality of multiple disorder predictors. Experiments on data used at the Critical Assessment of Techniques for Protein Structure Prediction (CASP7 and CASP8) have shown the effectiveness of the proposed algorithm.

Collaborators use vocabulary germane to the domain in context. Collaboration applications and collaborating systems use vocabularies at different (lower) layers that are specific to the state in which they execute. Distributed, yet collaborative domain-specific applications have demonstrated success

The proposed feature selection method aims to find a minimum subset of the most informative variables for classification/regression by efficiently approximating the Markov Blanket which is a set of variables that can shield a certain variable from the target. Instead of relying on the conditional independence test or network structure learning, the new method uses Hilbert-Schmidt Independence criterion as a measure of dependence among variables in a kernel-induced space. This allows effective approximation of the Markov Blanket that consists of multiple dependent features rather than being limited to a single feature. In addition, the new method can remove both irrelevant and redundant features at the same time. This method for discovering the Markov Blanket is applicable to both discrete and continuous variables, whereas previous methods cannot be used directly for continuous features and therefore are not applicable to regression problems. Experimental evaluations on synthetic and benchmark classification and regression datasets provide evidence that the new feature selection method can remove useless variables in low and in high dimensional problems more accurately than existing Markov Blanket based alternatives.

The analysis of social networks often assumes the time invariant scenario while in practice node attributes and links in such networks often evolve over time. In this paper, we propose a new method to predict node attributes and links in temporal networks.

Conditional random fields (CRF) are widely used for predicting output variables that have some internal structure. Most of the CRF research has been done on structured classification where the outputs are discrete. In this study we propose a CRF probabilistic model for structured regression that uses multiple non-structured predictors as its features. We construct features as squared prediction errors and show that this results in a Gaussian predictor. Learning becomes a convex optimization problem leading to a global solution for a set of parameters. Inference can be conveniently conducted through matrix computation. Experimental results on the remote sensing problem of estimating Aerosol Optical Depth (AOD) provide strong evidence that the proposed CRF model successfully exploits the inherent spatio-temporal properties of AOD data. The experiments revealed that CRF are more accurate than the baseline neural network and domain-based predictors.

In regression learning, it is often difficult to obtain the true values of the label variables, while multiple sources of noisy estimates of lower quality are readily available. To address this problem, we propose a new Bayesian approach that learns a regression model from data with noisy labels provided by multiple oracles. The proposed method provides closed form solution for model parameters and is applicable to both linear and nonlinear regression problems. In our experiments on synthetic and benchmark datasets this new regression model was consistently more accurate than a model trained with averaged estimates from multiple oracles as labels.

A typical approach in supervised learning is to select an accuracy measure and train a predictor that maximizes it. This can be insufficient in remote-sensing applications where predictor performance is often evaluated over multiple domain-specific accuracy measures. Here, we test the hypothesis that predictors can be trained to maximize performance over multiple accuracy measures. To do this, we evaluate several metalearning algorithms on the problem of aerosol optical depth (AOD) retrieval. The multiple accuracy measures included mean squared error, correlation, relative squared error, and fraction of satisfactory predictions. The proposed metalearning algorithms have a two-layer architecture, where the first layer consists of multiple neural networks, each trained using a different accuracy measure, and the second layer aggregates decisions of the first layer predictors. To evaluate AOD predictors, we used nearly 70 000 collocated data points whose attributes were radiances, solar and view angles, and terrain elevation from MODerate resolution Imaging Spectrometer (MODIS) instrument satellite observations and whose target AOD variable was obtained from the ground-based AEROsol robotic NETwork (AERONET) instruments. The data were collected at 221 AERONET locations over the globe in the period between 2005 and 2007. AOD prediction accuracies of neural networks were compared to the recently developed operational MODISC005retrieval algorithm and to several other data-mining methods. Results showed that neural networks are better at reproducing the test data than the operational retrieval algorithm and that predictors obtained by metalearning are robust over multiple accuracy measures.

In a data-mining approach, a model for estimation of aerosol optical depth (AOD) from satellite observations is learned using collocated satellite and ground-based observations. For accurate learning of such a spatio-temporal model, it is important to collect ground-based data from a large number of sites. The objective of this project is to determine appropriate locations for the next set of ground-based data collection sites to maximize accuracy of AOD estimation. Ideally, a new site should capture the most significant unseen aerosol patterns and should be the least correlated with the previously observed patterns. We propose achieving this aim by selecting the locations on which the existing prediction model is the most uncertain. Several criteria were considered for site selection, including uncertainty, spatial diversity, similarity in temporal pattern, and their combination. Extensive experiments on globally distributed data over 90 AERONET sites from the years 2005 and 2006 provide strong evidence that sites selected using the proposed algorithms improve the overall AOD prediction accuracy at a faster rate than those selected randomly or based on spatial diversity among sites.

BACKGROUND: Gene expression signatures in the mammalian brain hold the key to understanding neural development and neurological disease. Researchers have previously used voxelation in combination with microarrays for acquisition of genome-wide atlases of expression patterns in the mouse brain. On the other hand, some work has been performed on studying gene functions, without taking into account the location information of a gene's expression in a mouse brain. In this paper, we present an approach for identifying the relation between gene expression maps obtained by voxelation and gene functions. RESULTS: To analyze the dataset, we chose typical genes as queries and aimed at discovering similar gene groups. Gene similarity was determined by using the wavelet features extracted from the left and right hemispheres averaged gene expression maps, and by the Euclidean distance between each pair of feature vectors. We also performed a multiple clustering approach on the gene expression maps, combined with hierarchical clustering. Among each group of similar genes and clusters, the gene function similarity was measured by calculating the average gene function distances in the gene ontology structure. By applying our methodology to find similar genes to certain target genes we were able to improve our understanding of gene expression patterns and gene functions. By applying the clustering analysis method, we obtained significant clusters, which have both very similar gene expression maps and very similar gene functions respectively to their corresponding gene ontologies. The cellular component ontology resulted in prominent clusters expressed in cortex and corpus callosum. The molecular function ontology gave prominent clusters in cortex, corpus callosum and hypothalamus. The biological process ontology resulted in clusters in cortex, hypothalamus and choroid plexus. Clusters from all three ontologies combined were most prominently expressed in cortex and corpus callosum. CONCLUSION: The experimental results confirm the hypothesis that genes with similar gene expression maps might have similar gene functions. The voxelation data takes into account the location information of gene expression level in mouse brain, which is novel in related research. The proposed approach can potentially be used to predict gene functions and provide helpful suggestions to biologists.

In this paper we present an approach for identifying the relationships between gene expression maps and gene functions based on the multiplex gene expression maps of mouse brain obtained by voxelation. To analyze the dataset, we choose typical genes as queries and aim at discovering similar gene groups. We use the wavelet transform for extracting features from the left and right hemispheres averaged gene expression maps, and the Euclidean distance between each pair of feature vectors to determine gene similarity. We also perform a multiple clustering approach on the gene expression maps, combined with hierarchical clustering. Among each group of similar genes and clusters, the gene function similarity is measured by calculating the average gene function distances in the gene ontology structure. The experimental results confirm the hypothesis that genes with similar gene expression maps might have similar gene functions. The voxelation data takes into account the location information of gene expression level in mouse brain, which is novel in related research. The proposed approach can potentially be used to predict gene functions and provide helpful suggestions to biologists.

Identification of a small set of discriminative genes is a crucial step for effective prediction of disease or patient survival based on microarray gene expression data. Previous approaches to this problem are mainly based on analyzing differential gene expression data. In this work, an additional step is introduced to take advantage of prior knowledge about the relation of genes and a disease. In the proposed approach, keyword scanning of human proteins at the Swissprot database is performed to select genes related to the disease of interest followed by analysis of differential gene expressions. In results obtained on lung cancer data where a differential expression-based selection of genes is fairly inaccurate, our prior knowledge mining based approach offered a large improvement of prediction accuracy (0.74 vs. 0.58 ROC curve when using 20 genes). Furthermore, experimental results on a breast cancer dataset, where prediction based on differential gene expression alone was quite accurate can be further improved by integrating with our new approach.

One of the biggest challenges of current climate research is to characterize and quantify the effect of aerosols on the global and local weather. This requires an accurate prediction of aerosol optical density (AOD) which is defined as the amount of loss a beam of light incurs when it passes through the atmosphere. In this paper a neural network-based data-driven prediction model is considered which uses collocated satellite (MODIS) observation and ground-based (AERONET) AOD retrievals as predictors and target respectively. This paper studies an active learning-based data collection method which will facilitate the learning of a sufficiently accurate AOD prediction model using a minimal set of labeled training data by querying the labels of only the most informative data points.

In supervised learning, on data collected over space and time, different relationships can be found over different spatio-temporal regions. In such situations, an appropriate spatio-temporal data partitioning followed by building specialized predictors could often achieve higher overall prediction accuracy than when learning a single predictor on all the data. In practice, partitions are typically decided based on prior knowledge. As an alternative to domain-based partitioning, we propose a method that automatically discovers a spatio-temporal partitioning through the competition of regression models. The method is evaluated on a challenging problem using satellite observations to predict Aerosol Optical Depth (AOD), which represents the amount of depletion that a beam of radiation undergoes as it passes through the atmosphere. Our experiments used more than 20,000 labeled data points collected during 3 years from more than 100 sites worldwide. Our partitioning-based approach was compared to the recently developed operational AOD prediction algorithm, called C5, which uses domain knowledge for spatio-temporal partitioning of the Earth and implements a region-specific deterministic predictor that utilizes forward simulations from the postulated physical models. Data partitioning used in C5 divides the world into three spatio-temporal regions that differ based on the location and the time of the year as decided by domain experts. The results showed that a neural network predictor trained on all the data has accuracy comparable to C5. When specialized neural network predictors were learned on C5-based partitions, the overall prediction accuracy was not improved. On the other hand, our competition-based spatio-temporal data partitioning approach resulted in large accuracy improvements. The most accurate results were obtained when (1) the data from each of the sites were split into two temporal subsets, one for winter-spring months and another for summer-fall months; and (2) two neural network predictors were competing for each of the identified spatio-temporal subsets.

Aerosols are small airborne particles that both reflect and absorb incoming solar radiation and whose effect on the Earth's radiation budget is one of the biggest challenges of current climate research. To help address this challenge, numerous satellite sensors are employed to achieve global-scale monitoring of aerosols. Given the satellite measurements, the common objective is prediction of Aerosol Optical Depth (AOD). An important property of AOD is its low spatial variability on a scale of tens of kilometers. On the other hand, satellite sensors gather information in the form of multi-spectral images with high spatial resolution where pixels could be as small as a few hundred meters. Given an accurate ground-based AOD measurement over a specific location and time, all the pixels in the vicinity can be assumed to have the same AOD. If we treat satellite measurement at a single pixel as an instance, all pixels from the neighborhood can be considered as a bag of instances labeled with the same AOD. Given a number of bags obtained at numerous locations and at different times we can treat the problem of AOD prediction from satellite attributes as Multiple Instance Regression (MIR). An important challenge is that because of rapidly changing surface properties attribute values of pixels from a bag can vary a lot. This study evaluated several MIR approaches on several synthetic data sets and on a data set consisting of 800 labeled bags, each containing hundreds of pixel instances observed over the Continental U.S. by the MISR satellite instrument. The results indicate that the most successful MIR approach consists of an iterative procedure that detects and discards outlying instances and trains a predictor on the remaining ones.

Operational algorithms for retrieval of aerosols from satellite observations are typically created manually based on the domain knowledge. Validation studies, where the retrievals are compared to the available ground-truth data, are periodically performed with the goal of understanding how to further improve the quality of the retrieval algorithms. This letter describes a data-mining approach aimed to facilitate this highly labor-intensive process. It is based on training a neural network for retrieval and comparing its performance with that of the operational algorithm. The situations, where a neural network is more accurate, point to the weaknesses of the operational algorithm that could be corrected. Use of decision trees is proposed to provide easily interpretable descriptions of such situations. The approach was applied on 3646 collocated Moderate Resolution Imaging Spectroradiometer and AERONET observations over the continental U.S. related to the retrieval of aerosol optical thickness. The experiments showed that the approach is feasible and that it can be a valuable tool for the domain scientists working on the development of retrieval algorithms.

About 10 years ago we published our first predictor of intrinsically disordered protein residues in another IEEE journal, theProceedingsoftheIEEEInternationalConferenceonNeural Networks. Others call such proteins "natively unfolded" and "intrinsically unstructured." Since then, we and others have substantially improved the prediction of intrinsically disordered residues. The prediction of protein intrinsic disorder is similar to the prediction of secondary structure in terms of methodology, but, at the structural level, secondary structure (especially random coil) and intrinsic disorder differ completely in their dynamic motion. First, we will briefly describe the prediction of protein disorder, show the progress from ~ 70 % to ~ 85 % per residue prediction accuracy, and show that intrinsically disordered proteins are common over the three domains of life, but are especially common among the eukaryotes. Next we will discuss our methods for deducing functions that are associated with disordered rather than structured proteins. In brief, structured proteins have advantages for catalysis while disordered proteins and regions have advantages for the reversible, weak binding often observed in signaling, control, and regulation. After that we will discuss how disorder facilitates binding diversity in protein-protein interaction networks, both for single disordered regions binding to many partners and for many disordered regions with different sequences binding to a common site on the surface of one structured protein. Part three presents data indicating that alternative splicing is more prevalent in regions of RNA that code for disorder than those that code for structure, thus providing a means for evolving tissue-specific signaling networks. Finally, we will present a novel approach to drug discovery based on disordered protein.

Just over 10 years ago, in June, 1997, in the Proceedings of the IEEE International Conference on Neural Networks, we published our first predictor of intrinsically disordered protein [1]. Since then, we have substantially improved our predictors, and more than 20 other laboratory groups have joined in efforts to improve the prediction of protein disorder. At the algorithmic level, prediction of protein intrinsic disorder is similar to the prediction of secondary structure, but, at the structural level, secondary structure and intrinsic disorder are entirely different. The secondary structure class called random coil or irregular differs from intrinsic disorder due to very different dynamic properties, with the secondary structure class being much less mobile than the region of disorder. At the biological level, unlike the prediction of secondary structure, the prediction of intrinsic disorder has been revolutionary. That is, for many years, experimentalists have provided evidence that some proteins lack fixed structure or are disordered (or unfolded) under physiological conditions. Experimentalists further are showing that, for some proteins, functions depended on the unstructured rather than structured state. However, these examples have been mostly ignored. To our knowledge, not one disordered protein or disorder-associated function is discussed in any biochemistry textbook, even though such examples began to be discovered more than 50 years ago. Disorder prediction has been important for showing that the few experimentally characterized examples represent a very large cohort that is found all across all three domains of life.

Bioinformatics (2006) 22(23), 2876–2882

Abstract—This paper describes data mining approach for improving the accuracy of aerosol retrieval algorithms. The approach was applied on 1,722 collocated MODIS and AERONET observations over the western part of the continental U.S. Neural networks were trained to predict AERONET Aerosol Optical Thickness (AOT) using attributes derived from observations made by MODIS instrument onboard TERRA satellite. The results showed that neural networks provide more accurate retrievals than the operational MODIS algorithm. Study of differences between neural networks and the MODIS algorithm revealed useful information that can help domain scientists improve quality of the MODIS algorithm.

MOTIVATION: Attribute selection is a critical step in development of document classification systems. As a standard practice, words are stemmed and the most informative ones are used as attributes in classification. Owing to high complexity of biomedical terminology, general-purpose stemming algorithms are often conservative and could also remove informative stems. This can lead to accuracy reduction, especially when the number of labeled documents is small. To address this issue, we propose an algorithm that omits stemming and, instead, uses the most discriminative substrings as attributes. RESULTS: The approach was tested on five annotated sets of abstracts from iProLINK that report on the experimental evidence about five types of protein post-translational modifications. The experiments showed that Naive Bayes and support vector machine classifiers perform consistently better [with area under the ROC curve (AUC) accuracy in range 0.92-0.97] when using the proposed attribute selection than when using attributes obtained by the Porter stemmer algorithm (AUC in 0.86-0.93 range). The proposed approach is particularly useful when labeled datasets are small.

MOTIVATION: Modeling of protein interactions is often possible from known structures of related complexes. It is often time-consuming to find the most appropriate template. Hypothesized biological units (BUs) often differ from the asymmetric units and it is usually preferable to model from the BUs. RESULTS: ProtBuD is a database of BUs for all structures in the Protein Data Bank (PDB). We use both the PDBs BUs and those from the Protein Quaternary Server. ProtBuD is searchable by PDB entry, the Structural Classification of Proteins (SCOP) designation or pairs of SCOP designations. The database provides the asymmetric and BU contents of related proteins in the PDB as identified in SCOP and Position-Specific Iterated BLAST (PSI-BLAST). The asymmetric unit is different from PDB and/or Protein Quaternary Server (PQS) BUs for 52% of X-ray structures, and the PDB and PQS BUs disagree on 18% of entries. AVAILABILITY: The database is provided as a standalone program and a web server from http://dunbrack.fccc.edu/ProtBuD.php.

BACKGROUND: Due to the functional importance of intrinsically disordered proteins or protein regions, prediction of intrinsic protein disorder from amino acid sequence has become an area of active research as witnessed in the 6th experiment on Critical Assessment of Techniques for Protein Structure Prediction (CASP6). Since the initial work by Romero et al. (Identifying disordered regions in proteins from amino acid sequences, IEEE Int. Conf. Neural Netw., 1997), our group has developed several predictors optimized for long disordered regions (>30 residues) with prediction accuracy exceeding 85%. However, these predictors are less successful on short disordered regions (< or =30 residues). A probable cause is a length-dependent amino acid compositions and sequence properties of disordered regions. RESULTS: We proposed two new predictor models, VSL2-M1 and VSL2-M2, to address this length-dependency problem in prediction of intrinsic protein disorder. These two predictors are similar to the original VSL1 predictor used in the CASP6 experiment. In both models, two specialized predictors were first built and optimized for short (< or = 30 residues) and long disordered regions (>30 residues), respectively. A meta predictor was then trained to integrate the specialized predictors into the final predictor model. As the 10-fold cross-validation results showed, the VSL2 predictors achieved well-balanced prediction accuracies of 81% on both short and long disordered regions. Comparisons over the VSL2 training dataset via 10-fold cross-validation and a blind-test set of unrelated recent PDB chains indicated that VSL2 predictors were significantly more accurate than several existing predictors of intrinsic protein disorder. CONCLUSION: The VSL2 predictors are applicable to disordered regions of any length and can accurately identify the short disordered regions that are often misclassified by our previous disorder predictors. The success of the VSL2 predictors further confirmed the previously observed differences in amino acid compositions and sequence properties between short and long disordered regions, and justified our approaches for modelling short and long disordered regions separately. The VSL2 predictors are freely accessible for non-commercial use at http://www.ist.temple.edu/disprot/predictorVSL2.php.

In this study we proposed an iterative procedure for correcting sampling bias in labeled datasets for supervised learning applications. Given a much larger and unbiased unlabeled dataset, our approach relies on training contrast classifiers to iteratively select unlabeled examples most highly underrepresented in the labeled dataset. Once labeled, these examples could greatly reduce the sampling bias present in the labeled dataset. Unlike active learning methods, the actual labeling is not necessary in order to determine the most appropriate sampling schedule. The proposed procedure was applied on an important bioinformatics problem of prioritizing protein targets for structural genomics projects. We show that the procedure is capable of identifying protein targets that are underrepresented in current protein structure database, the Protein Data Bank (PDB). We argue that these proteins should be given higher priorities for experimental structural characterization to achieve faster sampling bias reduction in current PDB and make it more representative of the protein space.

UNLABELLED: The Database of Protein Disorder (DisProt) is a curated database that provides structure and function information about proteins that lack a fixed three-dimensional (3D) structure under putatively native conditions, either in their entirety or in part. Starting from the central premise that intrinsic disorder is an important structural class of protein and in order to meet the increasing interest thereof, DisProt is aimed at becoming a central repository of disorder-related information. For each disordered protein, the database includes the name of the protein, various aliases, accession codes, amino acid sequence, location of the disordered region(s), and methods used for structural (disorder) characterization. If applicable, most entries also list the biological function(s) of each disordered region, how each region of disorder is used for function, as well as provide links to PubMed abstracts and major protein databases. AVAILABILITY: www.disprot.org

We investigate the application of localized neural network-based distributional learning techniques for characterizing interesting groups and potentially new types of disorder proteins. Instead of employing a single autoassociator model for learning global distributions of ordered and disordered classes, clustering-based partitioning techniques are first applied independently to both ordered and disordered labeled data set to identify regions of similar characteristics. Subsequently, local autoassociators are employed on labeled data to learn distribution of each cluster. These local autoassociators are used in testing phase to assign each tuple from the unlabeled data set to the cluster closest in distributional sense. Obtained partitions are analyzed for the presence and frequency of the expert-annotated keywords. Frequency comparison is applied to provide insight of keywords sensitive to the distribution heterogeneity and disorder/order labeling. Experimental results on a labeled database of confirmed order and disorder proteins and unlabeled data extracted from SWISS/spl I.bar/PROT database are consistent with related literature and can provide further insight into relationship between protein similarity, keyword labeling and the disorder property.

Predictive data mining typically relies on labeled data without exploiting a much larger amount of available unlabeled data. We show that using unlabeled data can be beneficial in a range of important prediction problems and therefore should be an integral part of the learning process. Given an unlabeled dataset representative of the underlying distribution and a K-class labeled sample that might be biased, our approach is to learn K contrast classifiers each trained to discriminate a certain class of labeled data from the unlabeled population. We illustrate that contrast classifiers can be useful in one-class classification, outlier detection, density estimation, and learning from biased data. The advantages of the proposed approach are demonstrated by an extensive evaluation on synthetic data followed by real-life bioinformatics applications for (1) ranking PubMed articles by their relevance to protein disorder and (2) cost-effective enlargement of a disordered protein database.

A labeled sequence data set related to a certain biological property is often biased and, therefore, does not completely capture its diversity in nature. To reduce this sampling bias problem a data mining procedure is proposed for detecting underrepresented relevant sequences. The procedure is aimed at helping domain experts achieve a cost-effective qualitative enlargement of knowledge through an in-depth study of a small number of statistically underrepresented and functionally interesting sequences. Our procedure consists of: (i) learning a class-conditional distribution model on each class of labeled data; (ii) applying the models to select statistically underrepresented unlabeled sequences; and (iii) automatically evaluating their interestingness. An application of the proposed approach is illustrated on an important problem of increasing the data set of confirmed disordered proteins. The obtained results demonstrate the promise of the proposed approach for an efficient reduction of sampling bias in biological databases.

In this paper, we propose a general framework for distributed boosting intended for efficient integrating specialized classifiers learned over very large and distributed homogeneous databases that cannot be merged at a single location. Our distributed boosting algorithm can also be used as a parallel classification technique, where a massive database that cannot fit into main computer memory is partitioned into disjoint subsets for a more efficient analysis. In the proposed method, at each boosting round the classifiers are first learned from disjoint datasets and then exchanged amongst the sites. Finally the classifiers are combined into a weighted voting ensemble on each disjoint data set. The ensemble that is applied to an unseen test set represents an ensemble of ensembles built on all distributed sites. In experiments performed on four large data sets the proposed distributed boosting method achieved classification accuracy comparable or even slightly better than the standard boosting algorithm while requiring less memory and less computational time. In addition, the communication overhead of the distributed boosting algorithm is very small making it a viable alternative to the standard boosting for large-scale databases.

Combining multiple global models (e.g. back-propagation based neural networks) is an effective technique for improving classification accuracy. This technique reduces variance by manipulating the distribution of the training data. In many large scale data analysis problems involving heterogeneous databases with attribute instability, standard boosting methods can be improved by coalescing multiple classifiers. Each classifier uses different germane attribute information that is identified through the attribute selection process. We propose a new technique of boosting localized classifiers when heterogeneous data sets contain more homogeneous data distributions. Instead of a single global classifier for each boosting round, we have localized classifiers responsible for each homogeneous region. The number of regions is identified through a clustering algorithm performed at each boosting iteration. A new boosting method applied to real life spatial data and synthetic spatial data shows improvements in prediction accuracy when unstable driving attributes and heterogeneity are present in the data. In addition, boosting localized experts significantly reduces the number of iterations needed for achieving the maximal prediction accuracy.

1 Introduction Prediction of a continuous response variable in spatial-temporal domains has recently drawn attention in a data analysis community [4],[11]. Spatial-temporal regression models, learned on systematically collected values of driving attributes, can contribute to better understanding of complex phenomena studied in meteorology, oceanography, environmental science, precision agriculture and other domains. However, spatial-temporal modelling is often difficult due to various factors. For example, due to a prohibitive cost of data collection or other constraints, it is often not possible to systematically measure values of all attributes that have an influence to the observed response [15]. In such applications, models estimated on available attributes often have unsatisfactory explanatory power.

A boosting-based method for centers placement in radial basis function networks (RBFNs) is proposed. Also, the influence of several methods for drawing random samples on the accuracy of RBFNs is examined. The new method is compared to trivial, linear and non-linear regressors including the multilayer perceptron and alternative RBFN learning algorithms and its advantages are demonstrated for learning heterogeneous functions from sparse and non-uniform samples.

No abstract available.

Precision agriculture is a new approach to farming in which environmental characteristics at a sub-field level are used to guide crop production decisions. Instead of applying management actions and production inputs uniformly across entire fields, they are varied to match site-specific needs. A first step in this process is to define spatial regions having similar characteristics and to build local regression models describing the relationship between field characteristics and yield. From these yield prediction models, one can then determine optimum production input levels. Discovery of "similar" regions in fields is done by applying the DBSCAN clustering algorithm on data from more than one field, ignoring spatial attributes and the corresponding yield values. The experimental results on real life agriculture data show observable improvements in prediction accuracy, although there are many unresolved issues in applying the proposed method in practice.

For the data received online from a source with an unknown probability distribution, the question addressed in this article is how to efficiently partition it to smaller representative subsets (databases) and how to organize these data subsets in order to minimize the computational cost of the later data analysis. The proposed linear-time, online problem decomposition method achieves these objectives through balancing probability distributions of the individual disjoint data subsets, each aimed at approximating the original data-source distribution. Consequently, computationally efficient statistical data analysis and neural network modelling on data subsets fitting into a computer central memory will produce results similar to these obtained through a global, computationally infeasible data analysis. In addition, the proposed decomposition scheme enables for an effective distributed data analysis on a network of workstations.

A novel method for problem decomposition and for local model selection in a multimodel prediction system is proposed. The proposed method partitions the data into disjoint subsets obtained by the local regression modeling and then it learns the distributions on these sets in order to identify the most appropriate regression model for each test point. The system is applied to a site specific agriculture domain and is shown to provide a substantial improvement in the prediction quality as compared to a global model. Also, some aspects of local learner choice and setting of their parameters are discussed and an overall ability of the proposed model to accurately perform regression is assessed.

Precision agriculture data consisting of crop yield and topographic features are examined with the objective of explaining yield variability as a function of topographic attributes in order to extrapolate this knowledge to unseen agricultural sites. It is demonstrated that random data partitioning into training, validation and test subsets is not appropriate when dealing with agricultural problems characterized with strong spatial data correlation. A simple spatial data partitioning scheme that leads to significantly faster neural network training and slightly better generalization is proposed. Also, integration of predictors formed from spatially partitioned data led to improved generalization over a bagging integration procedure in experiments. The margin between the best spatial model and a trivial predictor for our precision agriculture problem was small indicating that topographic features alone could explain only a small amount of the yield variability.

The article summarizes the results of an exploratory data analysis on the INFFC competition time series. The analysis provides evidence that the problem is non-stationary and that the interpolation process for filling-in missing values alters the data distribution. The accuracy for trivial and linear predictors, determined in order to establish accuracy lower bounds for reasonable nonlinear prediction systems, identifies competition entries with prediction accuracies below the provided bounds. Finally, testbed design recommendations for future financial time series competitions are extracted from the results of this analysis.

A constructive learning algorithm dynamically creates a problem-specific neural network architecture rather than learning on a pre-specified architecture. The authors propose a parallel version of their recently presented constructive neural network learning algorithm. Parallelization provides a computational speedup by a factor of O(t) where t is the number of training examples. Distributed and parallel implementations under p4 using a network of workstations and a Touchstone DELTA are examined. Experimental results indicate that algorithm parallelization may result not only in improved computational time, but also in better prediction quality.>

For polynomially bounded weights and polylogarithmic precision, analog neural networks of polynomial size and depth 3 are strictly more powerful than those of polynomial size and depth 2.