Hasil untuk "cs.LG"

Naimur Rahman

Deep learning is often applied in settings where data are limited, correlated, and difficult to obtain, yet evaluation practices do not always reflect these constraints. Neuroimaging for prodromal Parkinsons disease is one such case, where subject numbers are small and individual scans produce many highly related samples. This work examines prodromal Parkinsons detection from resting-state fMRI as a machine learning problem centered on learning under extreme data scarcity. Using fMRI data from 40 subjects, including 20 prodromal Parkinsons cases and 20 healthy controls, ImageNet-pretrained convolutional neural networks are fine-tuned and evaluated under two different data partitioning strategies. Results show that commonly used image-level splits allow slices from the same subject to appear in both training and test sets, leading to severe information leakage and near-perfect accuracy. When a strict subject-level split is enforced, performance drops substantially, yielding test accuracies between 60 and 81 percent. Models with different capacity profiles are compared, including VGG19, Inception V3, Inception ResNet V2, and the lightweight MobileNet V1. Under subject-level evaluation, MobileNet demonstrates the most reliable generalization, outperforming deeper architectures despite having significantly fewer parameters. These results indicate that in extreme low-data regimes, evaluation strategy and model capacity have a greater impact on performance than architectural depth. Although the analysis is limited to a single cohort of 40 subjects and does not include external validation or cross-validation, it provides a concrete case study and practical recommendations for evaluating deep learning models under severe data scarcity.

Francisco Aguilera Moreno

Predicting cycling duration for a given route is essential for training planning and event preparation. Existing solutions rely on physics-based models that require extensive parameterization, including aerodynamic drag coefficients and real-time wind forecasts, parameters impractical for most amateur cyclists. This work presents a machine learning approach that predicts ride duration using route topology features combined with the athlete's current fitness state derived from training load metrics. The model learns athlete-specific performance patterns from historical data, substituting complex physical measurements with historical performance proxies. We evaluate the approach using a single-athlete dataset (N=96 rides) in an N-of-1 study design. After rigorous feature engineering to eliminate data leakage, we find that Lasso regression with Topology + Fitness features achieves MAE=6.60 minutes and R2=0.922. Notably, integrating fitness metrics (Chronic Training Load (CTL), Acute Training Load (ATL)) reduces error by 14% compared to topology alone (MAE=7.66 min), demonstrating that physiological state meaningfully constrains performance even in self-paced efforts. Progressive checkpoint predictions enable dynamic race planning as route difficulty becomes apparent.

Michel Youssef

We present a risk-calibrated approach to streaming intrusion detection that couples Bayesian Online Changepoint Detection (BOCPD) with decision thresholds aligned to Site Reliability Engineering (SRE) error budgets. BOCPD provides run-length posteriors that adapt to distribution shift and concept drift; we map these posteriors to alert decisions by optimizing expected operational cost under false-positive and false-negative budgets. We detail the hazard model, conjugate updates, and an O(1)-per-event implementation. A concrete SRE example shows how a 99.9% availability SLO (43.2 minutes per month error budget) yields a probability threshold near 0.91 when missed incidents are 10x more costly than false alarms. We evaluate on the full UNSW-NB15 and CIC-IDS2017 benchmarks with chronological splits, comparing against strong unsupervised baselines (ECOD, COPOD, and LOF). Metrics include PR-AUC, ROC-AUC, Brier score, calibration reliability diagrams, and detection latency measured in events. Results indicate improved precision-recall at mid to high recall and better probability calibration relative to baselines. We release implementation details, hyperparameters, and ablations for hazard sensitivity and computational footprint. Code and reproducibility materials will be made available upon publication; datasets and implementation are available from the corresponding author upon reasonable request.

Athanasios Angelakis, Amne Mousa, Micah L. A. Heldeweg et al.

Differentiating cardiogenic pulmonary oedema (CPE) from non-cardiogenic and structurally normal lungs in lung ultrasound (LUS) videos remains challenging due to the high visual variability of non-cardiogenic inflammatory patterns (NCIP/ARDS-like), interstitial lung disease, and healthy lungs. This heterogeneity complicates automated classification as overlapping B-lines and pleural artefacts are common. We introduce ZACH-ViT (Zero-token Adaptive Compact Hierarchical Vision Transformer), a 0.25 M-parameter Vision Transformer variant that removes both positional embeddings and the [CLS] token, making it fully permutation-invariant and suitable for unordered medical image data. To enhance generalization, we propose ShuffleStrides Data Augmentation (SSDA), which permutes probe-view sequences and frame orders while preserving anatomical validity. ZACH-ViT was evaluated on 380 LUS videos from 95 critically ill patients against nine state-of-the-art baselines. Despite the heterogeneity of the non-cardiogenic group, ZACH-ViT achieved the highest validation and test ROC-AUC (0.80 and 0.79) with balanced sensitivity (0.60) and specificity (0.91), while all competing models collapsed to trivial classification. It trains 1.35x faster than Minimal ViT (0.62M parameters) with 2.5x fewer parameters, supporting real-time clinical deployment. These results show that aligning architectural design with data structure can outperform scale in small-data medical imaging.

Michael Fuest, Alfredo Cuesta, Kalyan Veeramachaneni

Recent breakthroughs in large-scale generative modeling have demonstrated the potential of foundation models in domains such as natural language, computer vision, and protein structure prediction. However, their application in the energy and smart grid sector remains limited due to the scarcity and heterogeneity of high-quality data. In this work, we propose a method for creating high-fidelity electricity consumption time series data for rare and unseen context variables (e.g. location, building type, photovoltaics). Our approach, Context Encoding and Normalizing Time Series Generation, or CENTS, includes three key innovations: (i) A context normalization approach that enables inverse transformation for time series context variables unseen during training, (ii) a novel context encoder to condition any state-of-the-art time-series generator on arbitrary numbers and combinations of context variables, (iii) a framework for training this context encoder jointly with a time-series generator using an auxiliary context classification loss designed to increase expressivity of context embeddings and improve model performance. We further provide a comprehensive overview of different evaluation metrics for generative time series models. Our results highlight the efficacy of the proposed method in generating realistic household-level electricity consumption data, paving the way for training larger foundation models in the energy domain on synthetic as well as real-world data.

Timur Galimzyanov, Olga Kolomyttseva, Egor Bogomolov

We study retrieval design for code-focused generation tasks under realistic compute budgets. Using two complementary tasks from Long Code Arena -- code completion and bug localization -- we systematically compare retrieval configurations across various context window sizes along three axes: (i) chunking strategy, (ii) similarity scoring, and (iii) splitting granularity. (1) For PL-PL, sparse BM25 with word-level splitting is the most effective and practical, significantly outperforming dense alternatives while being an order of magnitude faster. (2) For NL-PL, proprietary dense encoders (Voyager-3 family) consistently beat sparse retrievers, however requiring 100x larger latency. (3) Optimal chunk size scales with available context: 32-64 line chunks work best at small budgets, and whole-file retrieval becomes competitive at 16000 tokens. (4) Simple line-based chunking matches syntax-aware splitting across budgets. (5) Retrieval latency varies by up to 200x across configurations; BPE-based splitting is needlessly slow, and BM25 + word splitting offers the best quality-latency trade-off. Thus, we provide evidence-based recommendations for implementing effective code-oriented RAG systems based on task requirements, model constraints, and computational efficiency.

Zahra Shahrooei, Ali Baheri

The primary goal of reinforcement learning is to develop decision-making policies that prioritize optimal performance, frequently without considering safety. In contrast, safe reinforcement learning seeks to reduce or avoid unsafe behavior. This paper views safety as taking actions with more predictable consequences under environment stochasticity and introduces a temporal difference algorithm that uses optimal transport theory to quantify the uncertainty associated with actions. By integrating this uncertainty score into the decision-making objective, the agent is encouraged to favor actions with more predictable outcomes. We theoretically prove that our algorithm leads to a reduction in the probability of visiting unsafe states. We evaluate the proposed algorithm on several case studies in the presence of various forms of environment uncertainty. The results demonstrate that our method not only provides safer behavior but also maintains the performance. A Python implementation of our algorithm is available at \href{https://github.com/SAILRIT/Risk-averse-TD-Learning}{https://github.com/SAILRIT/OT-guided-TD-Learning}.

Éloïse Benito-Rodriguez, Einar Urdshals, Jasmina Nasufi et al.

Understanding Large Language Models (LLMs) is key to ensure their safe and beneficial deployment. This task is complicated by the difficulty of interpretability of LLM structures, and the inability to have all their outputs human-evaluated. In this paper, we present the first step towards a predictive framework, where the genre of a text used to prompt an LLM, is predicted based on its activations. Using Mistral-7B and two datasets, we show that genre can be extracted with F1-scores of up to 98% and 71% using scikit-learn classifiers. Across both datasets, results consistently outperform the control task, providing a proof of concept that text genres can be inferred from LLMs with shallow learning models.

Haixia Liu

Neural collapse, a newly identified characteristic, describes a property of solutions during model training. In this paper, we explore neural collapse in the context of imbalanced data. We consider the $L$-extended unconstrained feature model with a bias term and provide a theoretical analysis of global minimizer. Our findings include: (1) Features within the same class converge to their class mean, similar to both the balanced case and the imbalanced case without bias. (2) The geometric structure is mainly on the left orthonormal transformation of the product of $L$ linear classifiers and the right transformation of the class-mean matrix. (3) Some rows of the left orthonormal transformation of the product of $L$ linear classifiers collapse to zeros and others are orthogonal, which relies on the singular values of $\hat Y=(I_K-1/N\mathbf{n}1^\top_K)D$, where $K$ is class size, $\mathbf{n}$ is the vector of sample size for each class, $D$ is the diagonal matrix whose diagonal entries are given by $\sqrt{\mathbf{n}}$. Similar results are for the columns of the right orthonormal transformation of the product of class-mean matrix and $D$. (4) The $i$-th row of the left orthonormal transformation of the product of $L$ linear classifiers aligns with the $i$-th column of the right orthonormal transformation of the product of class-mean matrix and $D$. (5) We provide the estimation of singular values about $\hat Y$. Our numerical experiments support these theoretical findings.

Singh Akansha

Graph Neural Networks (GNNs) revolutionize machine learning for graph-structured data, effectively capturing complex relationships. They disseminate information through interconnected nodes, but long-range interactions face challenges known as "over-squashing". This survey delves into the challenge of over-squashing in Graph Neural Networks (GNNs), where long-range information dissemination is hindered, impacting tasks reliant on intricate long-distance interactions. It comprehensively explores the causes, consequences, and mitigation strategies for over-squashing. Various methodologies are reviewed, including graph rewiring, novel normalization, spectral analysis, and curvature-based strategies, with a focus on their trade-offs and effectiveness. The survey also discusses the interplay between over-squashing and other GNN limitations, such as over-smoothing, and provides a taxonomy of models designed to address these issues in node and graph-level tasks. Benchmark datasets for performance evaluation are also detailed, making this survey a valuable resource for researchers and practitioners in the GNN field.

Zichao Li, Li Liu, Zeyu Wang et al.

Adversarial training (AT) with samples generated by Fast Gradient Sign Method (FGSM), also known as FGSM-AT, is a computationally simple method to train robust networks. However, during its training procedure, an unstable mode of "catastrophic overfitting" has been identified in arXiv:2001.03994 [cs.LG], where the robust accuracy abruptly drops to zero within a single training step. Existing methods use gradient regularizers or random initialization tricks to attenuate this issue, whereas they either take high computational cost or lead to lower robust accuracy. In this work, we provide the first study, which thoroughly examines a collection of tricks from three perspectives: Data Initialization, Network Structure, and Optimization, to overcome the catastrophic overfitting in FGSM-AT. Surprisingly, we find that simple tricks, i.e., a) masking partial pixels (even without randomness), b) setting a large convolution stride and smooth activation functions, or c) regularizing the weights of the first convolutional layer, can effectively tackle the overfitting issue. Extensive results on a range of network architectures validate the effectiveness of each proposed trick, and the combinations of tricks are also investigated. For example, trained with PreActResNet-18 on CIFAR-10, our method attains 49.8% accuracy against PGD-50 attacker and 46.4% accuracy against AutoAttack, demonstrating that pure FGSM-AT is capable of enabling robust learners. The code and models are publicly available at https://github.com/UCSC-VLAA/Bag-of-Tricks-for-FGSM-AT.

Rafael S. Pereira, Alexis Joly, Patrick Valduriez et al.

Deep learning models have become increasingly useful in many different industries. On the domain of image classification, convolutional neural networks proved the ability to learn robust features for the closed set problem, as shown in many different datasets, such as MNIST FASHIONMNIST, CIFAR10, CIFAR100, and IMAGENET. These approaches use deep neural networks with dense layers with softmax activation functions in order to learn features that can separate classes in a latent space. However, this traditional approach is not useful for identifying classes unseen on the training set, known as the open set problem. A similar problem occurs in scenarios involving learning on small data. To tackle both problems, few-shot learning has been proposed. In particular, metric learning learns features that obey constraints of a metric distance in the latent space in order to perform classification. However, while this approach proves to be useful for the open set problem, current implementation requires pair-wise training, where both positive and negative examples of similar images are presented during the training phase, which limits the applicability of these approaches in large data or large class scenarios given the combinatorial nature of the possible inputs.In this paper, we present a constraint-based approach applied to the representations in the latent space under the normalized softmax loss, proposed by[18]. We experimentally validate the proposed approach for the classification of unseen classes on different datasets using both metric learning and the normalized softmax loss, on disjoint and joint scenarios. Our results show that not only our proposed strategy can be efficiently trained on larger set of classes, as it does not require pairwise learning, but also present better classification results than the metric learning strategies surpassing its accuracy by a significant margin.

Irene Unceta, Jordi Nin, Oriol Pujol

When deployed in the wild, machine learning models are usually confronted with data and requirements that constantly vary, either because of changes in the generating distribution or because external constraints change the environment where the model operates. To survive in such an ecosystem, machine learning models need to adapt to new conditions by evolving over time. The idea of model adaptability has been studied from different perspectives. In this paper, we propose a solution based on reusing the knowledge acquired by the already deployed machine learning models and leveraging it to train future generations. This is the idea behind differential replication of machine learning models.

Alessio Benavoli, Dario Azzimonti, Dario Piga

Gaussian processes (GPs) are distributions over functions, which provide a Bayesian nonparametric approach to regression and classification. In spite of their success, GPs have limited use in some applications, for example, in some cases a symmetric distribution with respect to its mean is an unreasonable model. This implies, for instance, that the mean and the median coincide, while the mean and median in an asymmetric (skewed) distribution can be different numbers. In this paper, we propose Skew-Gaussian processes (SkewGPs) as a non-parametric prior over functions. A SkewGP extends the multivariate Unified Skew-Normal distribution over finite dimensional vectors to a stochastic processes. The SkewGP class of distributions includes GPs and, therefore, SkewGPs inherit all good properties of GPs and increase their flexibility by allowing asymmetry in the probabilistic model. By exploiting the fact that SkewGP and probit likelihood are conjugate model, we derive closed form expressions for the marginal likelihood and predictive distribution of this new nonparametric classifier. We verify empirically that the proposed SkewGP classifier provides a better performance than a GP classifier based on either Laplace's method or Expectation Propagation.

Yusuf Yazici

Credit card fraud is an ongoing problem for almost all industries in the world, and it raises millions of dollars to the global economy each year. Therefore, there is a number of research either completed or proceeding in order to detect these kinds of frauds in the industry. These researches generally use rule-based or novel artificial intelligence approaches to find eligible solutions. The ultimate goal of this paper is to summarize state-of-the-art approaches to fraud detection using artificial intelligence and machine learning techniques. While summarizing, we will categorize the common problems such as imbalanced dataset, real time working scenarios, and feature engineering challenges that almost all research works encounter, and identify general approaches to solve them. The imbalanced dataset problem occurs because the number of legitimate transactions is much higher than the fraudulent ones whereas applying the right feature engineering is substantial as the features obtained from the industries are limited, and applying feature engineering methods and reforming the dataset is crucial. Also, adapting the detection system to real time scenarios is a challenge since the number of credit card transactions in a limited time period is very high. In addition, we will discuss how evaluation metrics and machine learning methods differentiate among each research.

Alessio Benavoli, Dario Azzimonti, Dario Piga

Skew-Gaussian processes (SkewGPs) extend the multivariate Unified Skew-Normal distributions over finite dimensional vectors to distribution over functions. SkewGPs are more general and flexible than Gaussian processes, as SkewGPs may also represent asymmetric distributions. In a recent contribution we showed that SkewGP and probit likelihood are conjugate, which allows us to compute the exact posterior for non-parametric binary classification and preference learning. In this paper, we generalize previous results and we prove that SkewGP is conjugate with both the normal and affine probit likelihood, and more in general, with their product. This allows us to (i) handle classification, preference, numeric and ordinal regression, and mixed problems in a unified framework; (ii) derive closed-form expression for the corresponding posterior distributions. We show empirically that the proposed framework based on SkewGP provides better performance than Gaussian processes in active learning and Bayesian (constrained) optimization. These two tasks are fundamental for design of experiments and in Data Science.

Peter Gmeiner

Inferring the causal direction and causal effect between two discrete random variables X and Y from a finite sample is often a crucial problem and a challenging task. However, if we have access to observational and interventional data, it is possible to solve that task. If X is causing Y, then it does not matter if we observe an effect in Y by observing changes in X or by intervening actively on X. This invariance principle creates a link between observational and interventional distributions in a higher dimensional probability space. We embed distributions that originate from samples of X and Y into that higher dimensional space such that the embedded distribution is closest to the distributions that follow the invariance principle, with respect to the relative entropy. This allows us to calculate the best information-theoretic approximation for a given empirical distribution, that follows an assumed underlying causal model. We show that this information-theoretic approximation to causal models (IACM) can be done by solving a linear optimization problem. In particular, by approximating the empirical distribution to a monotonic causal model, we can calculate probabilities of causation. We can also use IACM for causal discovery problems in the bivariate, discrete case. However, experimental results on labeled synthetic data from additive noise models show that our causal discovery approach is lagging behind state-of-the-art approaches because the invariance principle encodes only a necessary condition for causal relations. Nevertheless, for synthetic multiplicative noise data and real-world data, our approach can compete in some cases with alternative methods.

Jonathan Klein, Sören Pirk, Dominik L. Michels

We present a novel domain adaptation framework that uses morphologic segmentation to translate images from arbitrary input domains (real and synthetic) into a uniform output domain. Our framework is based on an established image-to-image translation pipeline that allows us to first transform the input image into a generalized representation that encodes morphology and semantics - the edge-plus-segmentation map (EPS) - which is then transformed into an output domain. Images transformed into the output domain are photo-realistic and free of artifacts that are commonly present across different real (e.g. lens flare, motion blur, etc.) and synthetic (e.g. unrealistic textures, simplified geometry, etc.) data sets. Our goal is to establish a preprocessing step that unifies data from multiple sources into a common representation that facilitates training downstream tasks in computer vision. This way, neural networks for existing tasks can be trained on a larger variety of training data, while they are also less affected by overfitting to specific data sets. We showcase the effectiveness of our approach by qualitatively and quantitatively evaluating our method on four data sets of simulated and real data of urban scenes. Additional results can be found on the project website available at http://jonathank.de/research/eps/ .

Natalia Díaz-Rodríguez, Vincenzo Lomonaco, David Filliat et al.

Continual learning consists of algorithms that learn from a stream of data/tasks continuously and adaptively thought time, enabling the incremental development of ever more complex knowledge and skills. The lack of consensus in evaluating continual learning algorithms and the almost exclusive focus on forgetting motivate us to propose a more comprehensive set of implementation independent metrics accounting for several factors we believe have practical implications worth considering in the deployment of real AI systems that learn continually: accuracy or performance over time, backward and forward knowledge transfer, memory overhead as well as computational efficiency. Drawing inspiration from the standard Multi-Attribute Value Theory (MAVT) we further propose to fuse these metrics into a single score for ranking purposes and we evaluate our proposal with five continual learning strategies on the iCIFAR-100 continual learning benchmark.

Andrei N. Soklakov

We introduce the principle of Occam's Razor in a form which can be used as a basis for economical formulations of physics. This allows us to explain the general structure of the Lagrangian for a composite physical system, as well as some other artificial postulates behind the variational formulations of physical laws. As an example, we derive Hamilton's principle of stationary action together with the Lagrangians for the cases of Newtonian mechanics, relativistic mechanics and a relativistic particle in an external gravitational field.