Shortcut learning undermines model generalization to out-of-distribution data. While the literature attributes shortcuts to biases in superficial features, we show that imbalances in the semantic distribution of sample embeddings induce spurious semantic correlations, compromising model robustness. To address this issue, we propose SCISSOR (Semantic Cluster Intervention for Suppressing ShORtcut), a Siamese network-based debiasing approach that remaps the semantic space by discouraging latent clusters exploited as shortcuts. Unlike prior data-debiasing approaches, SCISSOR eliminates the need for data augmentation and rewriting. We evaluate SCISSOR on 6 models across 4 benchmarks: Chest-XRay and Not-MNIST in computer vision, and GYAFC and Yelp in NLP tasks. Compared to several baselines, SCISSOR reports +5.3 absolute points in F1 score on GYAFC, +7.3 on Yelp, +7.7 on Chest-XRay, and +1 on Not-MNIST. SCISSOR is also highly advantageous for lightweight models with ~9.5% improvement on F1 for ViT on computer vision datasets and ~11.9% for BERT on NLP. Our study redefines the landscape of model generalization by addressing overlooked semantic biases, establishing SCISSOR as a foundational framework for mitigating shortcut learning and fostering more robust, bias-resistant AI systems.
We introduce an attention-based method that uses learned binary attention masks to ensure that only attended image regions influence the prediction. Context can strongly affect object perception, sometimes leading to biased representations, particularly when objects appear in out-of-distribution backgrounds. At the same time, many image-level object-centric tasks require identifying relevant regions, often requiring context. To address this conundrum, we propose a two-stage framework: stage 1 processes the full image to discover object parts and identify task-relevant regions, while stage 2 leverages input attention masking to restrict its receptive field to these regions, enabling a focused analysis while filtering out potentially spurious information. Both stages are trained jointly, allowing stage 2 to refine stage 1. Extensive experiments across diverse benchmarks demonstrate that our approach significantly improves robustness against spurious correlations and out-of-distribution backgrounds.
The COVID-19 pandemic, caused by SARS-CoV-2, highlighted the critical need for accurate prediction of disease severity to optimize healthcare resource allocation and patient management. The spike protein, which facilitates viral entry into host cells, exhibits high mutation rates, particularly in the receptor-binding domain, influencing viral pathogenicity. Artificial intelligence approaches, such as deep learning, offer promising solutions for leveraging genomic and clinical data to predict disease outcomes. Objective: This study aimed to develop a hybrid CNN-LSTM deep learning model to predict COVID-19 severity using spike protein sequences and associated clinical metadata from South American patients. Methods: We retrieved 9,570 spike protein sequences from the GISAID database, of which 3,467 met inclusion criteria after standardization. The dataset included 2,313 severe and 1,154 mild cases. A feature engineering pipeline extracted features from sequences, while demographic and clinical variables were one-hot encoded. A hybrid CNN-LSTM architecture was trained, combining CNN layers for local pattern extraction and an LSTM layer for long-term dependency modeling. Results: The model achieved an F1 score of 82.92%, ROC-AUC of 0.9084, precision of 83.56%, and recall of 82.85%, demonstrating robust classification performance. Training stabilized at 85% accuracy with minimal overfitting. The most prevalent lineages (P.1, AY.99.2) and clades (GR, GK) aligned with regional epidemiological trends, suggesting potential associations between viral genetics and clinical outcomes. Conclusion: The CNN-LSTM hybrid model effectively predicted COVID-19 severity using spike protein sequences and clinical data, highlighting the utility of AI in genomic surveillance and precision public health. Despite limitations, this approach provides a framework for early severity prediction in future outbreaks.
Reliably monitoring and recognizing maritime vessels based on acoustic signatures is complicated by the variability of different recording scenarios. A robust classification framework must be able to generalize across diverse acoustic environments and variable source-sensor distances. To this end, we present a deep learning model with robust performance across different recording scenarios. Using a trainable spectral front-end and temporal feature encoder to learn a Gabor filterbank, the model can dynamically emphasize different frequency components. Trained on the VTUAD hydrophone recordings from the Strait of Georgia, our model, CATFISH, achieves a state-of-the-art 96.63 % percent test accuracy across varying source-sensor distances, surpassing the previous benchmark by over 12 percentage points. We present the model, justify our architectural choices, analyze the learned Gabor filters, and perform ablation studies on sensor data fusion and attention-based pooling.
Effective out-of-distribution (OOD) detection is crucial for the safe deployment of machine learning models in real-world scenarios. However, recent work has shown that OOD detection methods are vulnerable to adversarial attacks, potentially leading to critical failures in high-stakes applications. This discovery has motivated work on robust OOD detection methods that are capable of maintaining performance under various attack settings. Prior approaches have made progress on this problem but face a number of limitations: often only exhibiting robustness to attacks on OOD data or failing to maintain strong clean performance. In this work, we adapt an existing robust classification framework, TRADES, extending it to the problem of robust OOD detection and discovering a novel objective function. Recognising the critical importance of a strong clean/robust trade-off for OOD detection, we introduce an additional loss term which boosts classification and detection performance. Our approach, called HALO (Helper-based AdversariaL OOD detection), surpasses existing methods and achieves state-of-the-art performance across a number of datasets and attack settings. Extensive experiments demonstrate an average AUROC improvement of 3.15 in clean settings and 7.07 under adversarial attacks when compared to the next best method. Furthermore, HALO exhibits resistance to transferred attacks, offers tuneable performance through hyperparameter selection, and is compatible with existing OOD detection frameworks out-of-the-box, leaving open the possibility of future performance gains. Code is available at: https://github.com/hugo0076/HALO
As machine learning models grow in complexity and increasingly rely on publicly sourced data, such as the human-annotated labels used in training large language models, they become more vulnerable to label poisoning attacks. These attacks, in which adversaries subtly alter the labels within a training dataset, can severely degrade model performance, posing significant risks in critical applications. In this paper, we propose FLORAL, a novel adversarial training defense strategy based on support vector machines (SVMs) to counter these threats. Utilizing a bilevel optimization framework, we cast the training process as a non-zero-sum Stackelberg game between an attacker, who strategically poisons critical training labels, and the model, which seeks to recover from such attacks. Our approach accommodates various model architectures and employs a projected gradient descent algorithm with kernel SVMs for adversarial training. We provide a theoretical analysis of our algorithm's convergence properties and empirically evaluate FLORAL's effectiveness across diverse classification tasks. Compared to robust baselines and foundation models such as RoBERTa, FLORAL consistently achieves higher robust accuracy under increasing attacker budgets. These results underscore the potential of FLORAL to enhance the resilience of machine learning models against label poisoning threats, thereby ensuring robust classification in adversarial settings.
We propose an enhanced zeroth-order stochastic Frank-Wolfe framework to address constrained finite-sum optimization problems, a structure prevalent in large-scale machine-learning applications. Our method introduces a novel double variance reduction framework that effectively reduces the gradient approximation variance induced by zeroth-order oracles and the stochastic sampling variance from finite-sum objectives. By leveraging this framework, our algorithm achieves significant improvements in query efficiency, making it particularly well-suited for high-dimensional optimization tasks. Specifically, for convex objectives, the algorithm achieves a query complexity of O(d \sqrt{n}/\epsilon ) to find an epsilon-suboptimal solution, where d is the dimensionality and n is the number of functions in the finite-sum objective. For non-convex objectives, it achieves a query complexity of O(d^{3/2}\sqrt{n}/\epsilon^2 ) without requiring the computation ofd partial derivatives at each iteration. These complexities are the best known among zeroth-order stochastic Frank-Wolfe algorithms that avoid explicit gradient calculations. Empirical experiments on convex and non-convex machine learning tasks, including sparse logistic regression, robust classification, and adversarial attacks on deep networks, validate the computational efficiency and scalability of our approach. Our algorithm demonstrates superior performance in both convergence rate and query complexity compared to existing methods.
The rapid progress in deep generative models has led to the creation of incredibly realistic synthetic images that are becoming increasingly difficult to distinguish from real-world data. The widespread use of Variational Models, Diffusion Models, and Generative Adversarial Networks has made it easier to generate convincing fake images and videos, which poses significant challenges for detecting and mitigating the spread of misinformation. As a result, developing effective methods for detecting AI-generated fakes has become a pressing concern. In our research, we propose HFMF, a comprehensive two-stage deepfake detection framework that leverages both hierarchical cross-modal feature fusion and multi-stream feature extraction to enhance detection performance against imagery produced by state-of-the-art generative AI models. The first component of our approach integrates vision Transformers and convolutional nets through a hierarchical feature fusion mechanism. The second component of our framework combines object-level information and a fine-tuned convolutional net model. We then fuse the outputs from both components via an ensemble deep neural net, enabling robust classification performances. We demonstrate that our architecture achieves superior performance across diverse dataset benchmarks while maintaining calibration and interoperability.
This study investigates the applicability of Singular Value Decomposition for the image classification of specific breeds of cats and dogs using fur color as the primary identifying feature. Sequential Quadratic Programming (SQP) is employed to construct optimally weighted templates. The proposed method achieves 69% accuracy using the Frobenius norm at rank 10. The results partially validate the assumption that dominant features, such as fur color, can be effectively captured through low-rank approximations. However, the accuracy suggests that additional features or methods may be required for more robust classification, highlighting the trade-off between simplicity and performance in resource-constrained environments.
In heliophysics research, predicting solar flares is crucial due to their potential to impact both space-based systems and Earth's infrastructure substantially. Magnetic field data from solar active regions, recorded by solar imaging observatories, are transformed into multivariate time series to enable solar flare prediction using temporal window-based analysis. In the realm of multivariate time series-driven solar flare prediction, addressing severe class imbalance with effective strategies for multivariate time series representation learning is key to developing robust predictive models. Traditional methods often struggle with overfitting to the majority class in prediction tasks where major solar flares are infrequent. This work presents EXCON, a contrastive representation learning framework designed to enhance classification performance amidst such imbalances. EXCON operates through four stages: obtaining core features from multivariate time series data; selecting distinctive contrastive representations for each class to maximize inter-class separation; training a temporal feature embedding module with a custom extreme reconstruction loss to minimize intra-class variation; and applying a classifier to the learned embeddings for robust classification. The proposed method leverages contrastive learning principles to map similar instances closer in the feature space while distancing dissimilar ones, a strategy not extensively explored in solar flare prediction tasks. This approach not only addresses class imbalance but also offers a versatile solution applicable to univariate and multivariate time series across binary and multiclass classification problems. Experimental results, including evaluations on the benchmark solar flare dataset and multiple time series archive datasets with binary and multiclass labels, demonstrate EXCON's efficacy in enhancing classification performance.
Deep learning models for image classification have become standard tools in recent years. A well known vulnerability of these models is their susceptibility to adversarial examples. These are generated by slightly altering an image of a certain class in a way that is imperceptible to humans but causes the model to classify it wrongly as another class. Many algorithms have been proposed to address this problem, falling generally into one of two categories: (i) building robust classifiers (ii) directly detecting attacked images. Despite the good performance of these detectors, we argue that in a white-box setting, where the attacker knows the configuration and weights of the network and the detector, they can overcome the detector by running many examples on a local copy, and sending only those that were not detected to the actual model. This problem is common in security applications where even a very good model is not sufficient to ensure safety. In this paper we propose to overcome this inherent limitation of any static defence with randomization. To do so, one must generate a very large family of detectors with consistent performance, and select one or more of them randomly for each input. For the individual detectors, we suggest the method of neural fingerprints. In the training phase, for each class we repeatedly sample a tiny random subset of neurons from certain layers of the network, and if their average is sufficiently different between clean and attacked images of the focal class they are considered a fingerprint and added to the detector bank. During test time, we sample fingerprints from the bank associated with the label predicted by the model, and detect attacks using a likelihood ratio test. We evaluate our detectors on ImageNet with different attack methods and model architectures, and show near-perfect detection with low rates of false detection.
Performative learning addresses the increasingly pervasive situations in which algorithmic decisions may induce changes in the data distribution as a consequence of their public deployment. We propose a novel view in which these performative effects are modelled as push-forward measures. This general framework encompasses existing models and enables novel performative gradient estimation methods, leading to more efficient and scalable learning strategies. For distribution shifts, unlike previous models which require full specification of the data distribution, we only assume knowledge of the shift operator that represents the performative changes. This approach can also be integrated into various change-of-variablebased models, such as VAEs or normalizing flows. Focusing on classification with a linear-in-parameters performative effect, we prove the convexity of the performative risk under a new set of assumptions. Notably, we do not limit the strength of performative effects but rather their direction, requiring only that classification becomes harder when deploying more accurate models. In this case, we also establish a connection with adversarially robust classification by reformulating the minimization of the performative risk as a min-max variational problem. Finally, we illustrate our approach on synthetic and real datasets.
Label noise remains a challenge for training robust classification models. Most methods for mitigating label noise have been benchmarked using primarily datasets with synthetic noise. While the need for datasets with realistic noise distribution has partially been addressed by web-scraped benchmarks such as WebVision and Clothing1M, those benchmarks are restricted to the computer vision domain. With the growing importance of Transformer-based models, it is crucial to establish text classification benchmarks for learning with noisy labels. In this paper, we present AlleNoise, a new curated text classification benchmark dataset with real-world instance-dependent label noise, containing over 500,000 examples across approximately 5,600 classes, complemented with a meaningful, hierarchical taxonomy of categories. The noise distribution comes from actual users of a major e-commerce marketplace, so it realistically reflects the semantics of human mistakes. In addition to the noisy labels, we provide human-verified clean labels, which help to get a deeper insight into the noise distribution, unlike web-scraped datasets typically used in the field. We demonstrate that a representative selection of established methods for learning with noisy labels is inadequate to handle such real-world noise. In addition, we show evidence that these algorithms do not alleviate excessive memorization. As such, with AlleNoise, we set the bar high for the development of label noise methods that can handle real-world label noise in text classification tasks. The code and dataset are available for download at https://github.com/allegro/AlleNoise.
Robust classification of the operational environment of wireless devices is becoming increasingly important for wireless network optimization, particularly in a shared spectrum environment. Distinguishing between indoor and outdoor devices can enhance reliability and improve coexistence with existing, outdoor, incumbents. For instance, the unlicensed but shared 6 GHz band (5.925 - 7.125 GHz) enables sharing by imposing lower transmit power for indoor unlicensed devices and a spectrum coordination requirement for outdoor devices. Further, indoor devices are prohibited from using battery power, external antennas, and weatherization to prevent outdoor operations. As these rules may be circumvented, we propose a robust indoor/outdoor classification method by leveraging the fact that the radio-frequency environment faced by a device are quite different indoors and outdoors. We first collect signal strength data from all cellular and Wi-Fi bands that can be received by a smartphone in various environments (indoor interior, indoor near windows, and outdoors), along with GPS accuracy, and then evaluate three machine learning (ML) methods: deep neural network (DNN), decision tree, and random forest to perform classification into these three categories. Our results indicate that the DNN model performs the best, particularly in minimizing the most important classification error, that of classifying outdoor devices as indoor interior devices.
This study investigates the effects of radio-opaque artefacts, such as skin markers, breast implants, and pacemakers, on mammography classification models. After manually annotating 22,012 mammograms from the publicly available EMBED dataset, a robust multi-label artefact detector was developed to identify five distinct artefact types (circular and triangular skin markers, breast implants, support devices and spot compression structures). Subsequent experiments on two clinically relevant tasks $-$ breast density assessment and cancer screening $-$ revealed that these artefacts can significantly affect model performance, alter classification thresholds, and distort output distributions. These findings underscore the importance of accurate automatic artefact detection for developing reliable and robust classification models in digital mammography. To facilitate future research our annotations, code, and model predictions are made publicly available.
Random vector functional link (RVFL), a variant of single-layer feedforward neural network (SLFN), has garnered significant attention due to its lower computational cost and robustness to overfitting. Despite its advantages, the RVFL network's reliance on the square error loss function makes it highly sensitive to outliers and noise, leading to degraded model performance in real-world applications. To remedy it, we propose the incorporation of the HawkEye loss (H-loss) function into the RVFL framework. The H-loss function features nice mathematical properties, including smoothness and boundedness, while simultaneously incorporating an insensitive zone. Each characteristic brings its own advantages: 1) Boundedness limits the impact of extreme errors, enhancing robustness against outliers; 2) Smoothness facilitates the use of gradient-based optimization algorithms, ensuring stable and efficient convergence; and 3) The insensitive zone mitigates the effect of minor discrepancies and noise. Leveraging the H-loss function, we embed it into the RVFL framework and develop a novel robust RVFL model termed H-RVFL. Notably, this work addresses a significant gap, as no bounded loss function has been incorporated into RVFL to date. The non-convex optimization of the proposed H-RVFL is effectively addressed by the Nesterov accelerated gradient (NAG) algorithm, whose computational complexity is also discussed. The proposed H-RVFL model's effectiveness is validated through extensive experiments on $40$ benchmark datasets from UCI and KEEL repositories, with and without label noise. The results highlight significant improvements in robustness and efficiency, establishing the H-RVFL model as a powerful tool for applications in noisy and outlier-prone environments.
Human Activity Recognition (HAR) plays a crucial role in various applications such as human-computer interaction and healthcare monitoring. However, challenges persist in HAR models due to the data distribution differences between training and real-world data distributions, particularly evident in cross-user scenarios. This paper introduces a novel framework, termed Diffusion-based Noise-centered Adversarial Learning Domain Adaptation (Diff-Noise-Adv-DA), designed to address these challenges by leveraging generative diffusion modeling and adversarial learning techniques. Traditional HAR models often struggle with the diversity of user behaviors and sensor data distributions. Diff-Noise-Adv-DA innovatively integrates the inherent noise within diffusion models, harnessing its latent information to enhance domain adaptation. Specifically, the framework transforms noise into a critical carrier of activity and domain class information, facilitating robust classification across different user domains. Experimental evaluations demonstrate the effectiveness of Diff-Noise-Adv-DA in improving HAR model performance across different users, surpassing traditional domain adaptation methods. The framework not only mitigates distribution mismatches but also enhances data quality through noise-based denoising techniques.
Hyperspectral image classification, a task that assigns pre-defined classes to each pixel in a hyperspectral image of remote sensing scenes, often faces challenges due to the neglect of correlations between spectrally similar pixels. This oversight can lead to inaccurate edge definitions and difficulties in managing minor spectral variations in contiguous areas. To address these issues, we introduce the novel Dual-stage Spectral Supertoken Classifier (DSTC), inspired by superpixel concepts. DSTC employs spectrum-derivative-based pixel clustering to group pixels with similar spectral characteristics into spectral supertokens. By projecting the classification of these tokens onto the image space, we achieve pixel-level results that maintain regional classification consistency and precise boundary. Moreover, recognizing the diversity within tokens, we propose a class-proportion-based soft label. This label adaptively assigns weights to different categories based on their prevalence, effectively managing data distribution imbalances and enhancing classification performance. Comprehensive experiments on WHU-OHS, IP, KSC, and UP datasets corroborate the robust classification capabilities of DSTC and the effectiveness of its individual components. Code will be publicly available at https://github.com/laprf/DSTC.
As the adoption of Artificial Intelligence (AI) models expands into critical real-world applications, ensuring the explainability of these models becomes paramount, particularly in sensitive fields such as medicine and finance. Linear Discriminant Analysis (LDA) remains a popular choice for classification due to its interpretable nature, derived from its capacity to model class distributions and enhance class separation through linear combinations of features. However, real-world datasets often suffer from incomplete data, posing substantial challenges for both classification accuracy and model interpretability. In this paper, we introduce a novel and robust classification method, termed Weighted missing Linear Discriminant Analysis (WLDA), which extends LDA to handle datasets with missing values without the need for imputation. Our approach innovatively incorporates a weight matrix that penalizes missing entries, thereby refining parameter estimation directly on incomplete data. This methodology not only preserves the interpretability of LDA but also significantly enhances classification performance in scenarios plagued by missing data. We conduct an in-depth theoretical analysis to establish the properties of WLDA and thoroughly evaluate its explainability. Experimental results across various datasets demonstrate that WLDA consistently outperforms traditional methods, especially in challenging environments where missing values are prevalent in both training and test datasets. This advancement provides a critical tool for improving classification accuracy and maintaining model transparency in the face of incomplete data.
The recent generative AI models' capability of creating realistic and human-like content is significantly transforming the ways in which people communicate, create and work. The appropriate use of generative AI models can benefit society, while their misuse poses threats to the society. However, the lack of aligned multimodal datasets has inhibited the development of effective and robust methods for detecting machine-generated content, particularly in triple-modality settings (e.g., text, image, and voice). In this paper, we introduce RU-AI, a new large-scale multimodal dataset for robust and efficient detection of machine-generated content in text, image and voice. Our dataset is constructed on the basis of three large publicly available datasets: Flickr8K, COCO and Places205, by adding their corresponding AI duplicates, resulting total of 1,475,370 data instances. In addition, we create a noise variant of each modality of the datasets aiming to analyse the models' robustness. Given our dataset, we conduct extensive experiments with the current SOTA detection methods. The results reveal that existing models still struggle to achieve accurate and robust classification after training on our dataset. The RU-AI dataset is designed to support the development of detection methods across modalities and can be effectively utilised for identifying machine-generated content. The source code and dataset are available at https://github.com/ZhihaoZhang97/RU-AI.