Abstract Chest X-rays are an inexpensive and widely available imaging modality for diagnosing or monitoring a variety of medical conditions. Given their abundance, healthcare providers could greatly benefit from automated systems capable of screening healthy patients and supporting the diagnosis of pathological cases. Deep learning has become central to such decision-support systems, offering accurate and efficient image classification that can improve clinical workflows and reduce radiologist workload. However, despite the rapid evolution of general-purpose neural architectures, particularly attention-based models, their application to medical imaging remains constrained by limited incorporation of medical domain knowledge. Most existing attention mechanisms optimize only task-specific losses, disregarding crucial anatomical and lesion-location priors, which can hinder generalization and interpretability. In this work, we introduce a fully automated, attention-guided classification framework that integrates medical priors through an on-the-fly segmentation of the lungs, followed by a spatially aware attention loss that directs the network’s focus toward clinically relevant regions. The method requires minimal physician input—only a single annotated X-ray indicating potential lesion areas at initialization and generalizes effectively across patients without relying on absolute bounding-box coordinates. Gradient-based activation mapping is further employed to ensure alignment between attention and lesion-specific regions. Our approach is architecture-agnostic and integrates seamlessly into end-to-end pipelines. Experiments on two medical image datasets demonstrate that the proposed segmentation-enhanced attention loss improves both classification accuracy and representation interpretability compared to the standard cross-entropy loss. The code is available at: https://github.com/rcorizzo/cxr-segmentation-attention/ .

Dataset complexity poses a significant challenge in classification tasks, especially in real-world applications where a combination of factors such as class overlap, data imbalance, noise, and dimensionality can jeopardize a machine learning algorithm’s performance. While measures to quantify complexity have been proposed and studied in depth for tabular datasets, there is a lack of studies and toolkits focused on measuring complexity in non-structured image data. This limitation hinders our understanding of visual complexity, despite the importance of image data in fields such as healthcare, remote sensing, and autonomous navigation. To address this challenge, we introduce pycol-vis, a novel Python package that helps researchers estimate image complexity. The package implements 17 image complexity measures, specifically designed to capture complexity in real-world scenarios. This toolkit is essential for researchers dealing with complex classification problems in the vision domain, providing tools to assess difficulty and overlap in real-world image data.

Detecting multimodal misinformation on social media remains challenging due to inconsistencies between modalities, changes in temporal patterns, and substantial class imbalance. Many existing methods treat posts independently and fail to capture the event-level structure that connects them across time and modality. We propose E-CaTCH, an interpretable and scalable framework for robustly detecting misinformation. If needed, E-CaTCH clusters posts into pseudo-events based on textual similarity and temporal proximity, then processes each event independently. Within each event, textual and visual features are extracted using pre-trained BERT and ResNet encoders, refined via intra-modal self-attention, and aligned through bidirectional cross-modal attention. A soft gating mechanism fuses these representations to form contextualized, content-aware embeddings of each post. To model temporal evolution, E-CaTCH segments events into overlapping time windows and uses a trend-aware LSTM, enhanced with semantic shift and momentum signals, to encode narrative progression over time. Classification is performed at the event level, enabling better alignment with real-world misinformation dynamics. To address class imbalance and promote stable learning, the model integrates adaptive class weighting, temporal consistency regularization, and hard-example mining. The total loss is aggregated across all events. Extensive experiments on Fakeddit, IND, and COVID-19 MISINFOGRAPH demonstrate that E-CaTCH consistently outperforms state-of-the-art baselines. Cross-dataset evaluations further demonstrate its robustness, generalizability, and practical applicability across diverse misinformation scenarios.

Online hate speech proliferation has created a difficult problem for social media platforms. A particular challenge relates to the use of coded language by groups interested in both creating a sense of belonging for its users and evading detection. Coded language evolves quickly and its use varies over time. This paper proposes a methodology for detecting emerging coded hate-laden terminology. The methodology is tested in the context of online antisemitic discourse. The approach considers posts scraped from social media platforms, often used by extremist users. The posts are scraped using seed expressions related to previously known discourse of hatred towards Jews. The method begins by identifying the expressions most representative of each post and calculating their frequency in the whole corpus. It filters out grammatically incoherent expressions as well as previously encountered ones so as to focus on emergent well-formed terminology. This is followed by an assessment of semantic similarity to known antisemitic terminology using a fine-tuned large language model, and subsequent filtering out of the expressions that are too distant from known expressions of hatred. Emergent antisemitic expressions containing terms clearly relating to Jewish topics are then removed to return only coded expressions of hatred.

Deep learning for medical image classification is extremely important for decision support in medical healthcare settings. General-purpose neural network architectures for image classification have become increasingly sophisticated in recent years. Among them, attention-based models have provided significant advancements in deep learning. However, attention mechanisms adopted thus far focus on minimizing task-specific losses and do not fruitfully exploit medical knowledge, such as lesion-specific characteristics, during the training process, resulting in a potential reduction in accuracy. In this paper, we propose an attention-based approach that leverages knowledge of the localization of specific lesion types to guide the model training process. To this end, gradient-based activation mapping is used for incentivizing models to focus on the right area for a given lesion type. The approach is general since it can be applied to any deep learning architecture in end-to-end model training. Our experiments on two real-world medical image datasets show the ability of our approach to improve the classification performance of popular deep-learning model architectures over the classical cross-entropy loss.

Abstract Continual learning (CL) is one of the most promising trends in recent machine learning research. Its goal is to go beyond classical assumptions in machine learning and develop models and learning strategies that present high robustness in dynamic environments. This goal is realized by designing strategies that simultaneously foster the incorporation of new knowledge while avoiding forgetting past knowledge. The landscape of CL research is fragmented into several learning evaluation protocols, comprising different learning tasks, datasets, and evaluation metrics. Additionally, the benchmarks adopted so far are still distant from the complexity of real-world scenarios, and are usually tailored to highlight capabilities specific to certain strategies. In such a landscape, it is hard to clearly and objectively assess models and strategies. In this work, we fill this gap for CL on image data by introducing two novel CL benchmarks that involve multiple heterogeneous tasks from six image datasets, with varying levels of complexity and quality. Our aim is to fairly evaluate current state-of-the-art CL strategies on a common ground that is closer to complex real-world scenarios. We additionally structure our benchmarks so that tasks are presented in increasing and decreasing order of complexity—according to a curriculum—in order to evaluate if current CL models are able to exploit structure across tasks. We devote particular emphasis to providing the CL community with a rigorous and reproducible evaluation protocol for measuring the ability of a model to generalize and not to forget while learning. Furthermore, we provide an extensive experimental evaluation showing that popular CL strategies, when challenged with our proposed benchmarks, yield sub-par performance, high levels of forgetting, and present a limited ability to effectively leverage curriculum task ordering. We believe that these results highlight the need for rigorous comparisons in future CL works as well as pave the way to design new CL strategies that are able to deal with more complex scenarios.

A common challenge in classification and data mining is the class imbalance problem, where one class in a dataset has significantly more samples than another. The presence of this problem typically worsens the performance of classifiers trained on these datasets by giving them a strong preference for better-represented classes above others. Resampling is the process of adding (oversampling) or taking away (undersampling) samples from a dataset, typically with the purpose of introducing balance to the dataset before a classifier is trained. With recent advances in generative large language models (LLMs), certain authors have proposed the use of these models to generate new samples as a form of oversampling for imbalanced text data. This method has shown initial success, but has not been systematically compared to more traditional methods of resampling across many domains. We find that this use of generative LLMs generally outperforms other methods on imbalanced multiclass classification, but not on binary classification.

Abstract The massive adoption of social networks increased the need to analyze users’ data and interactions to detect and block the spread of propaganda and harassment behaviors, as well as to prevent actions influencing people towards illegal or immoral activities. In this paper, we propose HURI, a method for social network analysis that accurately classifies users assafeorrisky, according to their behavior in the social network. Specifically, the proposed hybrid approach leverages both the topology of the network of interactions and the semantics of the content shared by users, leading to an accurate classification also in the presence of noisy data, such as users who may appear to be risky due to the topic of their posts, but are actually safe according to their relationships. The strength of the proposed approach relies on the full and simultaneous exploitation of both aspects, giving each of them equal consideration during the combination phase. This characteristic makes HURI different from other approaches that fully consider only a single aspect and graft partial or superficial elements of the other into the first. The achieved performance in the analysis of a real-world Twitter dataset shows that the proposed method offers competitive performance with respect to eight state-of-the-art approaches.

Prolonged air leak is a complication arising from a collapsed lung which can lead to serious illness such as pneumonia and empyema, and patient suffering from indwelling chest tubes. Drainage of air and liquid from chest drains can be monitored and recorded using novel digital chest drainage devices. The collected data can be analyzed by predictive models, which can provide decision support in chest tube management. Despite the promising adoption of predictive models in this context, existing approaches are still in their infancy and are mostly based on autoregressive and conventional machine learning models. In this paper, we present a LSTM-based model architecture for air leak forecasting that is able to deal with non-linear dependencies among different features and contiguous time points. We devise a post-processing procedure that leverages predictions to suggest whether the patient could have their chest tube safely removed in the upcoming hours, and evaluate the results according to a medical protocol. Experimental results show that our model is able to outperform currently adopted models, in terms of both forecasting and classification performance, suggesting the feasibility of our approach for chest tube management.

Anomaly detection tools present the potential to enhance defense policies and protection against different types of threats, supporting public safety and national security. Lifelong anomaly detection showcases new and challenging scenarios in which models are challenged to automatically adapt to changing conditions without forgetting past knowledge. However, the presence of anomalies in incoming data may significantly impact the robustness of models in such scenarios. Although active learning strategies could be an asset to increase model longevity and robustness, they have never been explored in this context. In this paper, we propose an active lifelong anomaly detection framework for class-incremental scenarios that supports any memory-based experience replay method, any query strategy, and any anomaly detection model. While experience replay allows models to consolidate past knowledge and simultaneously adapt to new knowledge, an active learning module reduces the number of anomalies memorized in the replay buffer. We propose two strategies that automatically identify and remove additional data points that are likely to be anomalies based on the oracle’s feedback. Our experiments on popular host-based and network-based intrusion detection datasets show that our framework can improve the anomaly detection performance of models under low labeling budget constraints.

The detection of computer-generated text is an area of rapidly increasing significance as nascent generative models allow for efficient creation of compelling human-like text, which may be abused for the purposes of spam, disinformation, phishing, or online influence campaigns. Past work has studied detection of current state-of-the-art models, but despite a developing threat landscape, there has been minimal analysis of the robustness of detection methods to adversarial attacks. To this end, we evaluate neural and non-neural approaches on their ability to detect computer-generated text, their robustness against text adversarial attacks, and the impact that successful adversarial attacks have on human judgement of text quality. We find that while statistical features underperform neural features, statistical features provide additional adversarial robustness that can be leveraged in ensemble detection models. In the process, we find that previously effective complex phrasal features for detection of computer-generated text hold little predictive power against contemporary generative models, and identify promising statistical features to use instead. Finally, we pioneer the usage of $\Delta$MAUVE as a proxy measure for human judgement of adversarial text quality.

Change point detection methods offer a crucial ca-pability in modern data analysis tasks characterized by evolving time series data in the form of data streams. Recent interest in lifelong learning showed the importance of acquiring knowledge and identifying new occurring tasks in a continually evolving environment. Although this setting could benefit from a timely identification of changes, existing change point detection methods are unable to recognize recurring tasks, which is a necessary condition in lifelong learning. In this paper, we attempt to fill this gap by proposing LIFEWATCH, a novel Wasserstein-based change point detection approach with memory capable of modeling multiple data distributions in a fully unsupervised manner. Our method does not only detect changes, but discriminates between changes characterized by the appearance of a new task and changes that rather describe a recurring or previously seen task. An extensive experimental evaluation involving a large number of benchmark datasets shows that LIFEWATCH outperforms state-of-the-art methods for change detection while exploiting the characterization of detected changes to correctly identify tasks occurring in complex scenarios characterized by recurrence in lifelong consolidation settings.

Surgical risk assessments are central to the decision-making process utilized by medical professionals in the planning and execution of surgical procedures. In this regard, several tools are currently relied upon by surgeons to evaluate potential outcomes and associated risk. A popular and widely used assessment tool is the American College of Surgeons (ACS) National Surgical Quality Improvement Program (NSQIP) Universal Surgical Risk Calculator (ACS-SRC). One important limitation of this approach stands in its inability to exploit complex relationships among data variables, which may limit its classification accuracy. This paper tries to fill this gap by proposing a viable, low-resourced, and generalized decision support model using machine learning techniques that emulates the same general risk assessment functionality as the ACS-SRC, but utilizes more complex models such as gradient-boosted trees and deep neural networks. The experimental results show that our approach performs competitively with respect to state-of-the-art tools.

Recently, we have witnessed that deep learning-based approaches have been widely applied. Particularly, some applications involve data that are high dimensional, sparse or imbalanced, which are different from those applications with dense data processing, such as image classification and speech recognition, where deep learning-based approaches have been extensively studied. One of the main applications is the user-centric platform that consists of great deal of users, items and user generated tabular data which are quite high-dimensional. The characteristics of such data pose unique challenges to the adoption of deep learning in these applications, including modeling, training, and online serving, etc. More and more communities from both academia and industry have initiated the endeavors to solve these challenges. This workshop will provide a venue for both the research and engineering communities to discuss and formulate the challenges, utilize opportunities, and propose new ideas in the practice and theory of deep learning on high-dimensional, sparse and imbalanced data.

Recent advances in medical imaging and deep learning have enabled the efficient analysis of large databases of images. Notable examples include the analysis of computed tomography (CT), magnetic resonance imaging (MRI), and X-ray. While the automatic classification of images has proven successful, adopting such a paradigm in the medical healthcare setting is unfeasible. Indeed, the physician in charge of the detailed medical assessment and diagnosis of patients cannot trust a deep learning model’s decisions without further explanations or insights about their classification outcome. In this study, rather than relying on classification, we propose a new method that leverages deep neural networks to extract a representation of images and further analyze them through clustering, dimensionality reduction for visualization, and class activation mapping. Thus, the system does not make decisions on behalf of physicians. Instead, it helps them make a diagnosis. Experimental results on lung images affected by Pneumonia and Covid-19 lesions show the potential of our method as a tool for decision support in a medical setting. It allows the physician to identify groups of similar images and highlight regions of the input that the model deemed important for its predictions.

Detecting relevant changes in dynamic time series data in a timely manner is crucially important for many data analysis tasks in real-world settings. Change point detection methods have the ability to discover changes in an unsupervised fashion, which represents a desirable property in the analysis of unbounded and unlabeled data streams. However, one limitation of most of the existing approaches is represented by their limited ability to handle multivariate and high-dimensional data, which is frequently observed in modern applications such as traffic flow prediction, human activity recognition, and smart grids monitoring. In this paper, we attempt to fill this gap by proposing WATCH, a novel Wasserstein distance-based change point detection approach that models an initial distribution and monitors its behavior while processing new data points, providing accurate and robust detection of change points in dynamic high-dimensional data. An extensive experimental evaluation involving a large number of benchmark datasets shows that WATCH is capable of accurately identifying change points and outperforming state-of-the-art methods.

Structural concept complexity, class overlap, and data scarcity are some of the most important factors influencing the performance of classifiers under class imbalance conditions. When these effects were uncovered in the early 2000s, understandably, the classifiers on which they were demonstrated belonged to the classical rather than Deep Learning categories of approaches. As Deep Learning is gaining ground over classical machine learning and is beginning to be used in critical applied settings, it is important to assess systematically how well they respond to the kind of challenges their classical counterparts have struggled with in the past two decades. The purpose of this paper is to study the behavior of deep learning systems in settings that have previously been deemed challenging to classical machine learning systems to find out whether the depth of the systems is an asset in such settings. The results in both artificial and real-world image datasets show that these settings remain mostly challenging for Deep Learning systems. Deeper architectures help with structural concept complexity but not with data scarcity and class overlap.

Learning from imbalanced data poses significant challenges for the classifier. This becomes even more difficult, when dealing with multi-class problems. Here relationships among classes are no longer well-defined and it is easy to loose performance on one of the classes while gaining on other. In last years this topic has gained increased interest from the machine learning community - however, still there is a need for developing new and efficient algorithms to handle this challenge. In this paper we propose a new approach for balancing multi-class imbalanced problems. It is based on a two-step undersampling methodology. In the first step, a one-class classifier is being trained on each of the classes, achieving skew-insensitive data description. Support vectors for each class are extracted and used as new class representatives, thus achieving significant reduction in the terms of used instances. In the second step, an evolutionary undersampling approach is being used on these support vectors in order to further balance the training set. By applying this technique on a set of support vectors and not on a full dataset, we achieve a significant reduction of the computational time and increased accuracy. Finally, a standard multi-class classifier is being trained on the balanced data set. A thorough experimental study proves the usefulness of the proposed approach in comparison with state-of-the-art approaches for handling multi-class imbalanced data.

Recognizing human activities from multi-channel time series data collected from wearable sensors has become an important practical application of machine learning. A serious challenge comes from the presence of coherent activities or body movements, such as movements of the head while walking or sitting, since signals representing these movements are mixed and interfere with each other. Basic multi-label classification typically assumes independence within the multiple activities. This is oversimplified and reduces modeling power even when using state-of-the-art deep learning methods. In this paper, we investigate this new problem, which we name “Coherent Human Activity Recognition (Co-HAR)”, that keeps complete conditional dependency between the multiple labels. Additionally, we treat Co-HAR as a dense labelling problem that classifies each sample on a time step with multiple coherent labels to provide high-fidelity and duration-sensitive support to high-precision applications. To explicitly model conditional dependency, a novel condition-aware deep architecture “Conditional-UNet” is developed to allow for multiple dense labeling for Co-HAR. We also contribute a first-of-its-kind Co-HAR dataset for head gesture recognition associated with a user's activity, walking or sitting, to the research community. Extensive experiments on this dataset show that our model outperforms state-of-the-art deep learning methods and achieves up to 92% accuracy on context-based head gesture classification.

After lung surgery, a chest tube and a pump are used to manage air leaks and fluid drainage from the chest. The decision to remove or maintain the chest tube is based on drainage data collected from a digital pump that continuously monitors the patient. We construct a classifier to support this clinical decision-making process by identifying patients who may suffer adverse, extended air leaks early on. Intuitively, this problem can be modelled as a time-series fitted to monitoring data. However, we present a solution using a simple classifier constructed from data collected in a specific time frame (36- 48 hours) after surgery. We hypothesize that after surgery, patients struggle to attain a stable (favourable or adverse) status which prevails after a period of discrepancies and inconsistencies in the data. A solutions, we propose, is to identify this time frame when the majority of patients achieve their states of stability. Advantages of this approach include better classification performance with a lower burden of data collection during patient treatment. The paper presents the chest tube management as a classification task performed in a sliding window over time during patient monitoring. Our results show that reliable predictions can be achieved in the time window we identify, and that our classifier reduces unsafe chest tube removals at the expense of potentially maintaining a few that can be removed, i.e., we ensure that chest tubes that need to be maintained are not removed with potentially maintaining a few unnecessarily.

In many real-world applications, the characteristics of data collected by activity logs, sensors and mobile devices change over time. This behavior is known as concept drift. In complex environments, which produce high dimensional data streams, machine learning tasks become cumbersome, as models become outdated very quickly. In our study, we assess hundreds of combinations of data characteristics and methods on network traffic data. Specifically, we focus on seven conventional machine learning and deep learning methods and compare their generalization power in the presence of concept drift. Our results show that Convolutional Neural Networks (CNNs) outperform conventional methods, even when compared to an idealized upper bound on their performance created in a piecewise manner by selecting the best method and its best configuration at each point in time, thus mimicking the output of a perfect meta-learning architecture. In the context of sequential data subject to concept drift, our results appear to defy the usually accepted ”No Free Lunch Theorem (NFL)”, which stipulates that no method dominates all the others in every situation. While this is by no means a rejection of the NFL Theorem, which captures a much more complex phenomenon, it is nonetheless a surprising result worth further investigations. As a matter of fact, our results show that, when data availability is limited, a meta-learning approach is preferable to CNNs, as it requires less data for training.

Following a series of deep learning breakthroughs in the area of image segmentation, multiple objects in an image input can be finely sub-categorized. Although Convolutional Neural Networks (CNNs) are known for their state-of-the-art performance in image classification, they present drawbacks when used to analyze different data types, such as time series. In this paper, we propose the Sequential Mask Convolutional Neural Network (SMCNN), a method that overcomes such drawbacks, and leverages CNNs for sequential data analysis. Our method transforms sequential data into an image representation by means of a specialized filter that produces flexible shape forms, and detects multiple types of outliers simultaneously. We evaluate the effectiveness of our method on data containing a variety of anomaly types combined with different concept drifts. The solution shows to significantly outperform prior endeavors and to provide high generalization capabilities on a wide array of data characteristics. We attribute its success to its ability to pinpoint the exact location of patterns and anomalies in parallel and to the invariance of CNNs, which allows them to adapt seamlessly to concept drifts.

Machine learning based mobile traffic classification has become a popular topic in recent years. As mobile traffic data is dynamic in nature, the static model has become ineffective for the task of classifying future traffic. This is known as the concept drift problem in data streams. To this end, this paper presents an adaptive mobile traffic classification method. Specifically, a method based on the fuzzy competence model is devised to detect concept drift, and a dynamic learning method is presented to update the classification model, so as to adapt to an ever-changing environment at an appropriate time. The concept drift detection method relies on the data distribution instead of the classification error rate. Furthermore, the weights of flow samples are dynamically updated and flow samples are resampled for training a new model when a concept drift is detected. Moreover, recently trained models are saved and used for classification in weighted voting. The weight of each model is updated according to the performance it obtains on the most recent flow samples. On mobile traffic data, experimental results show that our proposed method obtains lower classification error rate with less time consumption on updating models as compared to related methods designed for handling concept drift problems.

As the distribution of a data stream evolves over time, a learner must adapt to its distributional shifts in order to make accurate predictions. In the context of anomaly detection, it is crucial for the learner to distinguish between natural changes in distribution and true anomalies in the data stream. This is the problem we focus on in this study which considers the situation where only normal data are available for initial training, but subsequent data can be either normal or anomalous. In that context, it is necessary to train a one-class learning anomaly detection system on the normal data and let the system output a score representing the degree of normalcy or outlierness that each data point in the subsequent data stream exhibits. The system then uses a threshold to discriminate between normal and anomalous instances. In the case of data stream, the data distribution may shift overtime, and a fixed threshold could develop a high false alarm rate or a low outlier detection rate in case of concept drift. To this end, we designed an adaptive sliding window approach which updates the threshold when necessary based on the scores distribution. Experimental results show that our method improves the performance of base anomaly detectors by dynamically updating the threshold of the scores when needed rather than using a fixed threshold or an adaptive threshold with fixed window sizes.

The class imbalance problem is a pervasive issue in many real-world domains. Oversampling methods that inflate the rare class by generating synthetic data are amongst the most popular techniques for resolving class imbalance. However, they concentrate on the characteristics of the minority class and use them to guide the oversampling process. By completely overlooking the majority class, they lose a global view on the classification problem and, while alleviating the class imbalance, may negatively impact learnability by generating borderline or overlapping instances. This becomes even more critical when facing extreme class imbalance, where the minority class is strongly underrepresented and on its own does not contain enough information to conduct the oversampling process. We propose a novel method for synthetic oversampling that uses the rich information inherent in the majority class to synthesize minority class data. This is done by generating synthetic data that is at the same Mahalanbois distance from the majority class as the known minority instances. We evaluate over 26 benchmark datasets, and show that our method offers a distinct performance improvement over the existing state-of-the-art in oversampling techniques.

Abstract Anomaly detection in streaming data is an important problem in numerous application domains. Most existing model‐based approaches to stream learning are based on decision trees due to their fast construction speed. This paper introduces streaming autoencoder (SA), a fast and novel anomaly detection algorithm based on ensembles of neural networks for evolving data streams. It is a one‐class learner, which only requires data from the positive class for training and is accurate even when anomalous training data are rare. It features an ensemble of threaded autoencoders with continuous learning capacity. Furthermore, the SA uses a 2‐step detection mechanism to ensure that real anomalies are detected with low false‐positive rates. The method is highly efficient because it processes data streams in parallel with multithreads and alternating buffers. Our analysis shows that SA has a linear runtime and requires constant memory space. Empirical comparisons to the state‐of‐the‐art methods on multiple benchmark data sets demonstrate that the proposed method detects anomalies efficiently with fewer false alarms.

Abstract In machine learning research and application, multiclass classification algorithms reign supreme. Their fundamental property is the reliance on the availability of data from all known categories to induce effective classifiers. Unfortunately, data from so‐called real‐world domains sometimes do not satisfy this property, and researchers use methods such as sampling to make the data more conducive for classification. However, there are scenarios in which even such explicit methods to rectify distributions fail. In such cases, 1‐class classification algorithms become the practical alternative. Unfortunately, domain complexity severely impacts their ability to produce effective classifiers. The work in this article addresses this issue and develops a strategy that allows for 1‐class classification over complex domains. In particular, we introduce the notion of learning along the lines of underlying domain concepts; an important source of complexity in domains is the presence of subconcepts, and by learning over them explicitly rather than on the entire domain as a whole, we can produce powerful 1‐class classification systems. The level of knowledge regarding these subconcepts will naturally vary by domain, and thus, we develop 3 distinct methodologies that take the amount of domain knowledge available into account. We demonstrate these over 3 real‐world domains.

Modern automobiles have been proven vulnerable to hacking by security researchers. By exploiting vulnerabilities in the car's external interfaces, such as wifi, bluetooth, and physical connections, they can access a car's controller area network (CAN) bus. On the CAN bus, commands can be sent to control the car, for example cutting the brakes or stopping the engine. While securing the car's interfaces to the outside world is an important part of mitigating this threat, the last line of defence is detecting malicious behaviour on the CAN bus. We propose an anomaly detector based on a Long Short-Term Memory neural network to detect CAN bus attacks. The detector works by learning to predict the next data word originating from each sender on the bus. Highly surprising bits in the actual next word are flagged as anomalies. We evaluate the detector by synthesizing anomalies with modified CAN bus data. The synthesized anomalies are designed to mimic attacks reported in the literature. We show that the detector can detect anomalies we synthesized with low false alarm rates. Additionally, the granularity of the bit predictions can provide forensic investigators clues as to the nature of flagged anomalies.

Multi-class learning is an important task in Data Science. One of the ways to achieve good performance on this task is to use Error Correcting Output Codes (ECOC), which is a powerful ensemble learning method that transforms a multi-class problem into a series of binary classifiers which it uses indirectly to learn the original multi-class problem. A crucial component of ECOC is the design of the coding matrix, which determines which binary problems should be combined to achieve multi-class classification. There are two general ways of designing the coding matrix. One is rooted in information theory while the other is data driven. In this work, we investigate the data-driven approach which was previously shown to bear greater promise and propose a better search through the coding-matrix space, keeping in mind the tradeoff between efficiency and effectiveness, as well as considerations about class-imbalance issues in the underlying binary problems. After consolidating our hy! potheses with a study on artificial domains, we propose the Unsupervised Deep Search Algorithm (UDS) coupled with re-sampling, to address both concerns. Our results on real world domains show that our method outperforms traditional multi-class learning methods.

Active learning is a common solution for reducing labeling costs and maximizing the impact of human labeling efforts in binary and multi-class classification settings. However, when we are faced with extreme levels of class imbalance, a situation in which it is not safe to assume that we have a representative sample of the minority class, it has been shown effective to replace the binary classifiers with a one-class classifiers. In such a setting, traditional active learning methods, and many previously proposed in the literature for one-class classifiers, prove to be inappropriate, as they rely on assumptions about the data that no longer stand. In this paper, we propose a novel approach to active learning designed for one-class classification. The proposed method does not rely on many of the inappropriate assumptions of its predecessors and leads to more robust classification performance. The gist of this method consists of labeling, in priority, the instances considered to fit the learned class the least by previous iterations of a one-class classification model. We provide empirical evidence for the merits of the proposed method compared to the available alternatives, and discuss how the method may have an impact in an applied setting.

This work examines the application of machine learning to an important area of medicine which aims to diagnose paediatric patients with β-thalassemia minor, iron deficiency anemia or the co-occurrence of these ailments. Iron deficiency anemia is a major cause of microcytic anemia and is considered an important task in global health. Whilst existing methods, based on linear equations, are proficient at distinguishing between the two classes of anemia, they fail to identify the co-occurrence of this issues. Machine learning algorithms, however, can induce non-linear decision boundaries that enable accurate classification within complex domains. Through a multi-label classification technique, known as problem transformations, we convert the learning task to one that is appropriate for machine learning and examine the effectiveness of machine learning algorithms on this domain. Our results show that machine learning classifiers produce good overall accuracy and are able to identify instances of the co-occurrence class unlike the existing methods.

Gamma-ray spectral classification requires the automatic identification of a large background class and a small minority class composed of instances that may pose a risk to humans and the environment. Accurate classification of such instances is required in a variety of domains, spanning event and port security to national monitoring for failures at industrial nuclear facilities. This work proposes a novel form of synthetic oversampling based on artificial neural network architecture and empirically demonstrates that it is superior to the state-of-the-art in synthetic oversampling on the target domain. In particular, we utilize gamma-ray spectral data collected for security purposes at the Vancouver 2010 winter Olympics and on a node of Health Canada's national monitoring networks.

MOTIVATION: Stem cell differentiation is largely guided by master transcriptional regulators, but it also depends on the expression of other types of genes, such as cell cycle genes, signaling genes, metabolic genes, trafficking genes, etc. Traditional approaches to understanding gene expression patterns across multiple conditions, such as principal components analysis or K-means clustering, can group cell types based on gene expression, but they do so without knowledge of the differentiation hierarchy. Hierarchical clustering can organize cell types into a tree, but in general this tree is different from the differentiation hierarchy itself. METHODS: Given the differentiation hierarchy and gene expression data at each node, we construct a weighted Euclidean distance metric such that the minimum spanning tree with respect to that metric is precisely the given differentiation hierarchy. We provide a set of linear constraints that are provably sufficient for the desired construction and a linear programming approach to identify sparse sets of weights, effectively identifying genes that are most relevant for discriminating different parts of the tree. RESULTS: We apply our method to microarray gene expression data describing 38 cell types in the hematopoiesis hierarchy, constructing a weighted Euclidean metric that uses just 175 genes. However, we find that there are many alternative sets of weights that satisfy the linear constraints. Thus, in the style of random-forest training, we also construct metrics based on random subsets of the genes and compare them to the metric of 175 genes. We then report on the selected genes and their biological functions. Our approach offers a new way to identify genes that may have important roles in stem cell differentiation. CONTACT: [email protected] SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Learning with class imbalance is always a challenging task in many real world applications such as the Internet, surveillance, security, and finance. Like many other successful machine learning algorithms, the success of the support vector machine (SVM) is limited when it is applied to the problem of learning from imbalanced datasets. SVM with different error costs has been widely used to deal with the class imbalanced problem. In this paper, we are trying to apply an instance-weighted variant of the SVM with both 1-norm and 2-norm format to deal with the class imbalance problem. We develop an asymmetric boosting method on the weights of the tradeoff parameters to optimize the instance-weighted SVM. The experimental results on the benchmark datasets show that the proposed algorithm is effective on the class imbalanced problem.

Multi-instance (MI) learning is different than standard propositional classification, as it uses a set of bags containing many instances as input. While the instances in each bag are not labeled, the bags themselves are, as positive or negative. In this paper, we present a novel multi-view, two-level classification framework to address the generalized multi-instance problems. We first apply supervised and unsupervised learning methods to transform a MI dataset into a multi-view, single meta-instance dataset. Then we develop a multi-view learning approach that can integrate the information acquired by individual view learners on the meta-instance dataset from the previous step, and construct a final model. Our empirical studies show that the proposed method performs well compared to other popular MI learning methods.

Tracking moving objects is a task of the utmost importance to the defence community. As this task requires high accuracy, rather than employing a single detector, it has become common to use multiple ones. In such cases, the tracks produced by these detectors need to be correlated (if they belong to the same sensing modality) or associated (if they were produced by different sensing modalities). In this work, we introduce Computational-Intelligence-based methods for correlating and associating various contacts and tracks pertaining to maritime vessels in an area of interest. Fuzzy k-Nearest Neighbours will be used to conduct track correlation and Fuzzy C-Means clustering will be applied for association. In that way, the uncertainty of the track correlation and association is handled through fuzzy logic. To better model the state of the moving target, the traditional Kalman Filter will be extended using an Echo State Network. Experimental results on five different types of sensing systems will be discussed to justify the choices made in the development of our approach. In particular, we will demonstrate the judiciousness of using Fuzzy k-Nearest Neighbours and Fuzzy C-Means on our tracking system and show how the extension of the traditional Kalman Filter by a recurrent neural network is superior to its extension by other methods.

Automatic Target Recognition (ATR) methods have been successfully applied to detect possible objects or regions of interest in sonar imagery. It is anticipated that the additional information obtained from additional views of an object should improve the classification performance over single-aspect classification. In this paper the detection of mine-like objects (MLO) on the seabed from multiple side-scan sonar views is considered. We transform the multiple-aspect classification problem into a multiple-instance learning problem and present a framework based upon the concepts of multiple-instance classifiers. Moreover, we present another framework based upon the Dempster-Shafer (DS) concept of fusion from single-view classifiers. Our experimental results indicate that both the presented frameworks can be successfully used in mine-like object classification.

Rapid detection of radioisotopes in gamma-ray data can, in some situations, be an important security concern. The task of designing an automated system for this purpose is complex due to, amongst other factors, the noisy nature of the data. The method described herein consists of preprocessing the data by applying a smoothing method tailored to gamma ray spectra, hoping that this should decrease their variance. Given that the number of counts at a given energy level in a spectrum should follow a Poisson distribution, smoothing may allow us to estimate the true photon arrival rate. Our experiments suggest that the added data preprocessing step can have large impact on the performance of anomaly detection algorithms on this particular domain.

In this paper, a novel learning method is introduced that borrows simultaneously from the principles of kernel methods and multi-layer perceptron. Specifically, the method implements the feature mapping idea of kernel methods into a multi-layer perceptron. Unlike in kernel learning where the feature space is usually invisible and inaccessible, the multilayer perceptron based mapping is explicit. Therefore, the proposed model can be learned directly in feature space. Together with the inherent sparse representation, the proposed approach will thus be much faster and easier to train even in the event of a large network size. The proposed approach is applied in the context of an Autonomous Underwater Vehicle Mine-Like Objects detection task. The results show that the proposed approach is able to improve upon the generalization performance of neural network based methods. Its prediction results are also close to or better than those obtained by kernel machines. Its learning and classification speed is shown to far surpass those of kernel machines. These results are confirmed on a number of experiments involving benchmarking UCI domains.

We have entered the big data age. Knowledge extraction from massive data is becoming more and more rewarding and urgent. MapReduce has provided a feasible framework for programming machine learning algorithms in Map and Reduce functions. The relatively simple programming interface has helped to solve machine learning algorithms' scalability problems. However, this framework suffers from an obvious weakness: it does not support iterations. This makes those algorithms requiring iterations difficult to fully explore the efficiency of MapReduce. In this paper, we propose to apply Meta-learning programmed with MapReduce to avoid parallelizing machine learning algorithms while also improving their scalability to big datasets. The experiments conducted on Hadoop fully distributed mode on Amazon EC2 demonstrate that our algorithm PML reduces the training computational complexity significantly when the number of computing nodes increases while gaining smaller error rates than those on one single node. The comparison of PML with the contemporary parallelized AdaBoost algorithm: AdaBoost.PL shows that PML has lower error rates.

With Canadian security and the safety of the general public in mind, physicists at Health Canada (HC) have begun to develop techniques to identify persons concealing radioactive material that may represent a threat to attendees at public gatherings, such as political proceedings and sporting events. To this end, Health Canada has initiated field trials that include the deployment of gamma-ray spectrometers. In particular, a series of these detectors, which take measurements every minute and produce 1,024 channel gamma-ray spectrum, were deployed during the Vancouver 2010 olympics. Simple computerized statistics and human expertise were used as the primary line of defence. More specifically, if a measured spectrum deviated significantly from the background, an internal alarm was sounded and an HC physicist undertook further analysis into the nature of the alarming spectrum. This strategy, however, lead to a significant number of costly and time consuming false positives. This research applies sophisticated machine learning algorithms to reduce the number of false positives to an acceptable level, the results of which are detailed in this paper. In addition, we emphasize the primary findings of our work and highlight avenues available to further improve upon our current results.

Binary classifiers have typically been the norm for building classification models in the Machine Learning community. However, an alternate to binary classification is one-class classification, which aims to build models using only a single class of data. This is particularly useful when there is an over-abundance of data of a particular class. In such imbalanced cases, binary classifiers may not perform very well, and one-class classifiers then become the viable option. In this paper, we are interested in investigating the performance of binary and one-class classifiers as the level of imbalance increases, and, thus, uncertainty in the second class. Our objective is to gain insight into which classification paradigm becomes more suitable as imbalance and uncertainty increase. To this end, we conduct experiments on various datasets, both artificial and from the UCI repository, and monitor the performance of the binary and one-class classifiers as the size of the second class gradually decreases, thus increasing the level of imbalance. The results show that as the level of imbalance increases, the performance of binary classifiers decreases, whereas one-class classifiers stay relatively stable.

Real world data mining applications such as Mine Countermeasure Missions (MCM) involve learning from imbalanced data sets, which contain very few instances of the minority classes and many instances of the majority class. For instance, the number of naturally occurring clutter objects (such as rocks) that are detected typically far outweighs the relatively rare event of detecting a mine. In this paper we propose support vector machine with adaptive asymmetric misclassification costs (instances weighted) to solve the skewed vector spaces problem in mine countermeasure missions. Experimental results show that the given algorithm could be used for imbalanced sonar image data sets and makes an improvement in prediction performance.

Verification of the Comprehensive Test-Ban-Treaty (CTBT), as a Pattern Recognition (PR) problem, has been proposed based on four radioxenon features. It has been noted, however, that in many cases this limited feature set is insufficient to distinguish radioxenon levels effected by an explosion from those that are solely products of industrial activities. As a means of improving the detectability of low-yield clandestine nuclear explosions, this paper motivates the inclusion of meteorological indicators in the CTBT feature-space, promotes further research into which meteorological indicators are most informative, and how they may be acquired. In doing so, we present classification results from four simulated scenarios. These results demonstrate that the inclusion of a simple wind direction feature can significantly increase the prospect of classifying challenging detonation events, and suggests the predictive power of meteorological features in general.

Algorithm performance evaluation is so entrenched in the machine learning community that one could call it an addiction. Like most addictions, it is harmful and very difficult to give up. It is harmful because it has serious limitations. Yet, we have great faith in practicing it in a ritualistic manner: we follow a fixed set of rules telling us the measure, the data sets and the statistical test to use. When we read a paper, even as reviewers, we are not sufficiently critical of results that follow these rules. Here, we will debate what are the limitations and how to best address them. This article may not cure the addiction but hopefully it will be a good first step along that road.

No abstract available.

Recently, progressive border sampling (PBS) was proposed for sample selection in supervised learning by progressively learning an augmented full border from small labeled datasets. However, this quadratic learning algorithm is inapplicable to large datasets. In this paper, we incorporate the PBS to a state of the art technique called coupling Markov chain Monte Carlo (CMCMC) in an attempt to scale the original algorithm up on large labeled datasets. The CMCMC can produce an exact sample while a naive strategy for Markov chain Monte Carlo cannot guarantee the convergence to a stationary distribution. The resulting CMCMC-PBS algorithm is thus proposed for border sampling on large datasets. CMCMC-PBS exhibits several remarkable characteristics: linear time complexity, learner-independence, and a consistent convergence to an optimal sample from the original training sets by learning from their subsamples. Our experimental results on the 33 either small or large labeled datasets from the UCIKDD repository and a nuclear security application show that our new approach outperforms many previous sampling techniques for sample selection.

Classifier performance evaluation typically gives rise to vast numbers of results that are difficult to interpret. On the one hand, a variety of different performance metrics can be applied; and on the other hand, evaluation must be conducted on multiple domains to get a clear view of the classifier's general behaviour. In this paper, we present a visualization technique that allows a user to study the results from a domain point of view and from a classifier point of view. We argue that classifier evaluation should be done on an exploratory basis. In particular, we suggest that, rather than pre-selecting a few metrics and domains to conduct our evaluation on, we should use as many metrics and domains as possible and mine the results of this study to draw valid and relevant knowledge about the behaviour of our algorithms. The technique presented in this paper will enable such a process.

Anti-virus systems traditionally use signatures to detect malicious executables, but signatures are over-fitted features that are of little use in machine learning. Other more heuristic methods seek to utilize more general features, with some degree of success. In this paper, we present a data mining approach that conducts an exhaustive feature search on a set of computer viruses and strives to obviate over-fitting. We also evaluate the predictive power of a classifier by taking into account dependence relationships that exist between viruses, and we show that our classifier yields high detection rates and can be expected to perform as well in real-world conditions.

This research introduces a new paradigm: dynamic document organization, DDO, for managing documents. DDO consists of managing documents automatically under the assumption that the topic structure in a collection of documents is always variable and temporary. This paradigm is in contrast with static document organization, SDO, the currently used paradigm which assumes that the topic structure is fixed and permanent. In this work, we consider two scenarios, a three-phase-scenario and a four-phase-scenario, for managing documents based on DDO. In both scenarios, text clustering, cluster identification, and document classification are integrated into a cycle. In the four-phase-scenario, one more phase, classifier training, is added between the cluster identification and document classification phases. The goal of this research is to evaluate the two proposed scenarios and contrast the best one ti its best SDO counterpart. We show that the four-phase DDO scenario is more reliable than the three-phase DDO scenario, and that it generally outperforms the best SDO scenario.

Text clustering is the process of segmenting a particular collection of texts into subgroups including content-based similar ones. This study proposes a new neural network, called NTSO (neural text self organizer), which is suitable for text clustering. This neural network uses string vectors instead of numerical vectors as its input vectors and its weight vectors are different from those of other unsupervised neural networks such as Kohonen networks and ART (adaptive resonance theory), although it is similar to Kohonen networks at the architecture level and in its learning process. Intuitively, text is better represented by a string vector than by a numerical vector. The representation of texts into numerical vectors leads to two main problems: sparse distribution and huge dimensionality of the feature vectors. This study proposes an unsupervised neural network that uses string vectors for text clustering, to address these problems.

Resampling methods are commonly used for dealing with the class‐imbalance problem. Their advantage over other methods is that they are external and thus, easily transportable. Although such approaches can be very simple to implement, tuning them most effectively is not an easy task. In particular, it is unclear whether oversampling is more effective than undersampling and which oversampling or undersampling rate should be used. This paper presents an experimental study of these questions and concludes that combining different expressions of the resampling approach is an effective solution to the tuning problem. The proposed combination scheme is evaluated on imbalanced subsets of the Reuters‐21578 text collection and is shown to be quite effective for these problems.

We introduce a new learning algorithm for decision lists to allow features that are constructed from the data and to allow a trade- ofi between accuracy and complexity. We bound its generalization error in terms of the number of errors and the size of the classifler it flnds on the training data. We also compare its performance on some natural data sets with the set covering machine and the support vector machine.

A common misperception within the neural network community is that even with nonlinearities in their hidden layer, autoassociators trained with backpropagation are equivalent to linear methods such as principal component analysis (PCA). Our purpose is to demonstrate that nonlinear autoassociators actually behave differently from linear methods and that they can outperform these methods when used for latent extraction, projection, and classification. While linear autoassociators emulate PCA, and thus exhibit a flat or unimodal reconstruction error surface, autoassociators with nonlinearities in their hidden layer learn domains by building error reconstruction surfaces that, depending on the task, contain multiple local valleys. This interpolation bias allows nonlinear autoassociators to represent appropriate classifications of nonlinear multimodal domains, in contrast to linear autoassociators, which are inappropriate for such tasks. In fact, autoassociators with hidden unit nonlinearities can be shown to perform nonlinear classification and nonlinear recognition.

Possible paradigms for concept learning by feedforward neural networks include discrimination and recognition. An interesting aspect of this dichotomy is that the recognition-based implementation can learn certain domains much more efficiently than the discrimination-based one, despite the close structural relationship between the two systems. The purpose of this paper is to explain this difference in efficiency. We suggest that it is caused by a difference in the generalization strategy adopted by the backpropagation procedure in both cases: while the autoassociator uses a (fast) bottom-up strategy, MLP has recourse to a (slow) top-down one, despite the fact that the two systems are both optimized by the backpropagation procedure. This result is important because it sheds some light on the nature of backpropagation's adaptive capability. From a practical viewpoint, it suggests a deterministic way to increase the efficiency of backpropagation-trained feedforward networks.

Machine translation of locative prepositions is not straightforward, even between closely related languages. This paper discusses a system of translation of locative prepositions between English and French. The system is based on the premises that English and French do not always conceptualize objects in the same way, and that this accounts for the major differences in the ways that locative prepositions are used in these languages. This paper introduces knowledge representations of conceptualizations of objects, and a method for translating prepositions based on these conceptual representations.