Chaotic dynamics are ubiquitous in many real-world systems, ranging from biological and industrial processes to climate dynamics and the spread of viruses. These systems are characterized by high sensitivity to initial conditions, making it challenging to predict their future behavior confidently. In this study, we propose a novel deep-learning framework that addresses this challenge by directly exploiting the long-term compounding of local prediction errors during model training, aiming to extend the time horizon for reliable predictions of chaotic systems. Our approach observes the future trajectories of initial errors at a time horizon, modeling the evolution of the loss to that point through the use of two major components: (1) a recurrent architecture (Error Trajectory Tracing) designed to trace the trajectories of predictive errors through phase space, and (2) a training regime, Horizon Forcing, that pushes the model’s focus out to a predetermined time horizon. We validate our method on three classic chaotic systems and six real-world time series prediction tasks with chaotic characteristics. The results show that our approach outperforms the state-of-the-art methods.

Large Language Models (LLMs) have shown strong general capabilities in many applications. However, how to make them reliable tools for some specific tasks such as automated short answer grading (ASAG) remains a challenge. We present SteLLA (Structured Grading System Using LLMs with RAG) in which a) Retrieval Augmented Generation (RAG) approach is used to empower LLMs specifically on the ASAG task by extracting structured information from the highly relevant and reliable external knowledge based on the instructor-provided reference answer and rubric, b) an LLM performs a structured and question-answering-based evaluation of student answers to provide analytical grades and feedback. A real-world dataset which contains students’ answers in an exam was collected from a college-level Biology course. Experiments show that our proposed system can achieve substantial agreement with the human grader while providing break-down grades and feedback on all the knowledge points examined in the problem. A qualitative and error analysis of the feedback generated by GPT4 shows that GPT4 is good at capturing facts while may prone to inferring too much implication from the given text in the grading task which provides insights into the usage of LLMs in the ASAG system.

Analyzing animal behavior from video data is crucial for understanding brain function, assessing pharmacological interventions, and examining genetic modifications. Traditional methods often struggle to accurately analyze group behaviors in complex environments. To address these challenges, we introduce the Animal Joint Embedded Prediction Architecture (Animal-JEPA), a novel self-supervised learning model designed for studying animal behavior from video data. Animal-JEPA leverages a dynamic scaling mechanism and an elliptical masking strategy to enhance feature extraction and behavioral analysis without the need for labeled data. Our approach significantly outperforms existing models, including Separate 3D ConvNet (S3D) [3], Video Vision Transformer (ViViT) [4], and the original V-JEPA [6], particularly in multi-category and multi-objective classification tasks on our newly developed Mice-Behavior3 (MB3) dataset. The results highlight Animal-JEPA’s potential to improve the accuracy and adaptability of behavioral analysis in animal research, providing a powerful tool for neuroscientists and researchers.

Effective early warning systems for extreme flood events in large river basins necessitate reliable long-lead streamflow forecasts. However, the inherent uncertainty within each phase of the weather system-rainfall prediction, runoff generation, and streamflow prediction-amplifies with each stage, rendering accurate long-lead streamflow estimations challenging. In response to this, our study introduces a novel deep-learning-based model, the Cascaded Spatio-Temporal Learning Deep Network (CASTLE). CASTLE synergistically integrates observed upstream precipitation, recent streamflow data, and short-term precipitation forecasts derived from a selection of quantitative climate models to produce an accurate streamflow estimate. Specifically, we employ deep residual architectures on both observed and forecasted precipitation data to model the cascading spatio-temporal processes, which begin with upstream rainfall, move to rainfall-runoff, and finally conclude with downstream discharge. Our aim is to identify hidden space-time patterns that can be used to forecast future downstream flow over extended periods. We assess CASTLE’s efficacy by forecasting the downstream discharge of the Ganges River over a long lead time. Results show that our approach outperforms the current state-of-the-art streamflow forecasting models.

The understanding of chaotic systems is challenging not only for theoretical research but also for many important applications. Chaotic behavior is found in many nonlinear dynamical systems, such as those found in climate dynamics, weather, the stock market, and the space-time dynamics of virus spread. A reliable solution for these systems must handle their complex space-time dynamics and sensitive dependence on initial conditions. We develop a deep learning framework to push the time horizon at which reliable predictions can be made further into the future by better evaluating the consequences of local errors when modeling nonlinear systems. Our approach observes the future trajectories of initial errors at a time horizon to model the evolution of the loss to that point with two major components: 1) a recurrent architecture, Error Trajectory Tracing, that is designed to trace the trajectories of predictive errors through phase space, and 2) a training regime, Horizon Forcing, that pushes the model’s focus out to a predetermined time horizon. We validate our method on classic chaotic systems and real-world time series prediction tasks with chaotic characteristics, and show that our approach outperforms the current state-of-the-art methods.

The study of model bias and variance with respect to decision boundaries is critically important in supervised learning and artificial intelligence. There is generally a tradeoff between the two, as fine-tuning of the decision boundary of a classification model to accommodate more boundary training samples (i.e., higher model complexity) may improve training accuracy (i.e., lower bias) but hurt generalization against unseen data (i.e., higher variance). By focusing on just classification boundary fine-tuning and model complexity, it is difficult to reduce both bias and variance. To overcome this dilemma, we take a different perspective and investigate a new approach to handle inaccuracy and uncertainty in the training data labels, which are inevitable in many applications where labels are conceptual entities and labeling is performed by human annotators. The process of classification can be undermined by uncertainty in the labels of the training data; extending a boundary to accommodate an inaccurately labeled point will increase both bias and variance. Our novel method can reduce both bias and variance by estimating the pointwise label uncertainty of the training set and accordingly adjusting the training sample weights such that those samples with high uncertainty are weighted down and those with low uncertainty are weighted up. In this way, uncertain samples have a smaller contribution to the objective function of the model’s learning algorithm and exert less pull on the decision boundary. In a real-world physical activity recognition case study, the data present many labeling challenges, and we show that this new approach improves model performance and reduces model variance.

Clustering is an important unsupervised learning method with serious challenges when data is sparse and high-dimensional. Generated clusters are often evaluated with general measures, which may not be meaningful or useful for practical applications and domains. Using a distance metric, a clustering algorithm searches through the data space, groups close items into one cluster, and assigns far away samples to different clusters. In many real-world applications, the number of dimensions is high and data space becomes very sparse. Selection of a suitable distance metric is very difficult and becomes even harder when categorical data is involved. Moreover, existing distance metrics are mostly generic, and clusters created based on them will not necessarily make sense to domain-specific applications. One option to address these challenges is to integrate domain-defined rules and guidelines into the clustering process. In this work we propose a GAN-based approach called Catalysis Clustering to incorporate domain knowledge into the clustering process. With GANs we generate catalysts, which are special synthetic points drawn from the original data distribution and verified to improve clustering quality when measured by a domain-specific metric. We then perform clustering analysis using both catalysts and real data. Final clusters are produced after catalyst points are removed. Experiments on two challenging real-world datasets clearly show that our approach is effective and can generate clusters that are meaningful and useful for real-world applications.

A real-world system often exhibits complex dynamics arising from interaction among its subunits. In machine learning and data mining, these interactions are usually formulated as dependency and correlation among system variables. Similar to Convolution Neural Network dealing with spatially correlated features and Recurrent Neural Network with temporally correlated features, in this paper we present a novel deep learning model to tackle functionally interactive features by stacking a Conditional Restricted Boltzmann Machine and a Deep Neural Network (CRBM-DNN). Variables with their dependency relationships are organized into a bipartite graph, which is further converted into a Restricted Boltzmann Machine conditioned by domain knowledge. We integrate this CRBM and a DNN into one deep learning model constrained by one overall cost function. CRBM-DNN can solve both supervised and unsupervised learning problems. Compared to a regular neural network of the same size, CRBM-DNN has fewer parameters so they require fewer training samples. We perform extensive comparative studies with a large number of supervised learning and unsupervised learning methods using several challenging real-world datasets, and achieve significant superior performance.

More and more user comments like Tweets are available, which often contain user concerns. In order to meet the demands of users, a good summary generating from multiple documents should consider reader interests as reflected in reader comments. In this article, we focus on how to generate a summary from multi-document documents by considering reader comments, named as reader-aware multi-document summarization (RA-MDS). We present an innovative topic-based method for RA-MDA, which exploits latent topics to obtain the most salient and lessen redundancy summary from multiple documents. Since finding latent topics for RA-MDS is a crucial step, we also present a Heterogeneous-length Text Topic Modeling (HTTM) to extract topics from the corpus that includes both news reports and user comments, denoted as heterogeneous-length texts. In this case, the latent topics extract by HTTM cover not only important aspects of the event, but also aspects that attract reader interests. Comparisons on summary benchmark datasets also confirm that the proposed RA-MDS method is effective in improving the quality of extracted summaries. In addition, experimental results demonstrate that the proposed topic modeling method outperforms existing topic modeling algorithms.

This paper presents a new algorithm, Reinforced and Informed Network-based Clustering(RINC), for finding unknown groups of similar data objects in sparse and largely non-overlapping feature space where a network structure among features can be observed. Sparse and non-overlapping unlabeled data become increasingly common and available especially in text mining and biomedical data mining. RINC inserts a domain informed model into a modelless neural network. In particular, our approach integrates physically meaningful feature dependencies into the neural network architecture and soft computational constraint. Our learning algorithm efficiently clusters sparse data through integrated smoothing and sparse auto-encoder learning. The informed design requires fewer samples for training and at least part of the model becomes explainable. The architecture of the reinforced network layers smooths sparse data over the network dependency in the feature space. Most importantly, through back-propagation, the weights of the reinforced smoothing layers are simultaneously constrained by the remaining sparse auto-encoder layers that set the target values to be equal to the raw inputs. Empirical results demonstrate that RINC achieves improved accuracy and renders physically meaningful clustering results.

We propose a new approach, CCRBoost, to identify the hierarchical structure of spatio-temporal patterns at different resolution levels and subsequently construct a predictive model based on the identified structure. To accomplish this, we first obtain indicators within different spatio-temporal spaces from the raw data. A distributed spatio-temporal pattern (DSTP) is extracted from a distribution, which consists of the locations with similar indicators from the same time period, generated by multi-clustering. Next, we use a greedy searching and pruning algorithm to combine the DSTPs in order to form an ensemble spatio-temporal pattern (ESTP). An ESTP can represent the spatio-temporal pattern of various regularities or a non-stationary pattern. To consider all the possible scenarios of a real-world ST pattern, we then build a model with layers of weighted ESTPs. By evaluating all the indicators of one location, this model can predict whether a target event will occur at this location. In the case study of predicting crime events, our results indicate that the predictive model can achieve 80 percent accuracy in predicting residential burglary, which is better than other methods.

The development of disastrous flood forecasting techniques able to provide warnings at a long lead-time (5-15 days) is of great importance to society. Extreme Flood is usually a consequence of a sequence of precipitation events occurring over from several days to several weeks. Though precise short-term forecasting the magnitude and extent of individual precipitation event is still beyond our reach, long-term forecasting of precipitation clusters can be attempted by identifying persistent atmospheric regimes that are conducive for the precipitation clusters. However, such forecasting will suffer from overwhelming number of relevant features and high imbalance of sample sets. In this paper, we propose an integrated data mining framework for identifying the precursors to precipitation event clusters and use this information to predict extended periods of extreme precipitation and subsequent floods. We synthesize a representative feature set that describes the atmosphere motion, and apply a streaming feature selection algorithm to online identify the precipitation precursors from the enormous feature space. A hierarchical re-sampling approach is embedded in the framework to deal with the imbalance problem. An extensive empirical study is conducted on historical precipitation and associated flood data collected in the State of Iowa. Utilizing our framework a few physically meaningful precipitation cluster precursor sets are identified from millions of features. More than 90% of extreme precipitation events are captured by the proposed prediction model using precipitation cluster precursors with a lead time of more than 5 days.

High dimensional data visualization is critical to data analysts since it gives a direct view of original data. We present a method to visualize large amount of high dimensional data. We divide dimensions of data into several groups. Then, we use one icon to represent each group, and associate visual properties of each icon with dimensions in each group. A high dimensional data record will be represented by multiple different types of icons located in the same position. Furthermore, we use summary icons to display local details of viewer's interests and the whole data set at meantime. We show its effectiveness and efficiency through a case study on a real large data set.

Clustering is a widely used knowledge discovery technique. It helps uncovering structures in data that were not previously known. The clustering of large data sets has received a lot of attention in recent years, however, clustering is a still a challenging task since many published algorithms fail to do well in scaling with the size of the data set and the number of dimensions that describe the points, or in finding arbitrary shapes of clusters, or dealing effectively with the presence of noise. In this paper, we present a new clustering algorithm, based in the fractal properties of the data sets. The new algorithm, which we call Fractal Clustering (FC), places points incrementally in the cluster for which the change in the fractal dimension after adding the point is the least. This is a very natural way of clustering points, since points in the same cluster have a great degree of self-similarity among them (and much less self-similarity with respect to points in other clusters). FC requires one scan of the data, is suspendable at will, providing the best answer possible at that point, and is incremental. We show via experiments that FC effectively deals with large data sets, high-dimensionality and noise and is capable of recognizing clusters of arbitrary shape.