Counterfactual explanations have been proposed to provide actionable insights for complex black-box classifiers by suggesting “what-if” explanations that alter undesirable prediction outcomes to desired ones. Recent advances have emphasized desirable properties such as causality, plausibility, and contextual alignment; however, existing methods often address these objectives in isolation. In particular, causal counterfactual approaches enforce structural dependencies among features, typically assuming access to a fully specified structural causal model, while context-aware counterfactual frameworks focus on improving alignment with user-defined contextual constraints. As a result, a gap remains between context-aware and causally grounded counterfactual generation, especially in latent-space representations, which can jointly address multiple objectives and improve validity across domains. In this work, we introduce CausalCACTUS, a causally consistent and context-aware counterfactual explanation framework that extends the CACTUS method by integrating automatically learned causal structure into latent-space counterfactual generation using a composite -VAE model. CausalCACTUS incorporates a directed acyclic graph learned from data-driven causal discovery to impose causal feasibility constraints, together with a phase-based latent optimization strategy that improves convergence stability and counterfactual validity while preserving user-defined contextual constraints. Our experimental evaluation across five publicly available tabular datasets and multiple baseline models demonstrates that CausalCACTUS achieves improved context alignment and causal consistency metrics while maintaining competitive performance on standard counterfactual metrics, such as validity and proximity. Ablation studies and qualitative examples further illustrate that jointly enforcing context preservation and causal structure leads to more actionable, trustworthy counterfactual explanations that respect user-defined contexts and causal dependencies in real-world scenarios.

Publicly available phonocardiogram (PCG) datasetsremain limited in size and pathological diversity, constrainingboth auscultation training and the generalisation of automatedheart-sound classifiers. A class-conditional diffusion model forPCG generation is developed in the log-mel domain and synthetic fidelity is assessed using complementary (i) physiologyinspired plausibility metrics, (ii) downstream label-consistencyevaluation, and (iii) expert listening. Experiments use the PhysioNet/Computing in Cardiology Challenge 2016 dataset (3240recordings) with recording-level splits. After preprocessing andquality control, 16,749 non-overlapping 4 s clips are mapped to anormalised 1×128×128 log-mel representation to train a conditional 2D U-Net denoiser with classifier-free guidance. Signal-levelplausibility is quantified on reconstructed waveforms using threelightweight metrics: an envelope-autocorrelation rhythm score,an amplitude-based explosion score, and the dominant cyclelag. Synthetic clips preserve similar dominant cycle durationsbut exhibit reduced envelope periodicity and increased transientburstiness relative to real clips. For downstream evaluation, aResNet-50 classifier achieves 92.24% accuracy on the held-outreal test set and 82.8% accuracy on class-balanced syntheticbatches, indicating that generated signals retain discriminativestructure relevant to normal/abnormal classification. In a pilotexpert listening study (60 clips, two clinicians), most syntheticclips are judged as heart-sound-like, while abnormality sensitivityis low for both real and synthetic 4 s excerpts. Overall, the resultsprovide a practical baseline for diffusion-based PCG generationwhile highlighting remaining challenges in retaining abnormalacoustic cues and reducing reconstruction-induced artefacts.

Cognitive impairment progressively limits independent living, yet scalable methods for identifying individuals at future risk remain limited. In the 2026 George B. Moody PhysioNet Challenge, our team (Matcha) developed a hierarchical framework for predicting cognitive impairment from overnight polysomnography (PSG). Modality-specific convolutional stems learned a 192- dimensional representation for every 30-s epoch from human and automated sleep annotations together with crossview self-supervision. Local and full-night Transformers then aggregated the epoch sequence into a patientlevel prediction, while a low-capacity record-wise residual represented whole-night sleep summaries and outcomeobservation timing. In the three-site leave-one-site-out (LOSO) validation, it achieved a macro age-conditioned AUROC of 0.804. On the official phase hidden validation set, the best Small- and Large-track entries achieved 0.847 and 0.844, respectively, ranking 1st out of 513 entries.

We consider data-driven Bayesian state estimation from compressed measurements (BSCM) of a model-free process. The dimension of the temporal measurement vector is lower than that of the temporal state vector to be estimated, leading to an under-determined inverse problem. The underlying dynamical model of the state’s evolution is unknown for a ‘model-free process.’ Hence, it is difficult to use traditional model-driven methods, for example, Kalman and particle filters. Instead, we consider data-driven methods. We experimentally show that two existing unsupervised learning-based data-driven methods fail to address the BSCM problem in a model-free process. The methods are-data-driven nonlinear state estimation (DANSE) and deep Markov model (DMM). While DANSE provides good predictive/ forecasting performance to model the temporal measurement data as a time series, its unsupervised learning lacks suitable regularization for tackling the BSCM task. We then propose a semi-supervised learning approach and develop a semi-supervised learning-based DANSE method, referred to as SemiDANSE. In SemiDANSE, we use a large amount of unlabelled data along with a limited amount of labelled data, i.e., pairwise measurement-and-state data, which provides the desired regularization. Using benchmark chaotic dynamical systems, we empirically show that the data-driven SemiDANSE provides competitive state estimation performance for BSCM using a handful of different measurement systems, against a hybrid method called KalmanNet and two model-driven methods (extended Kalman filter and unscented Kalman filter) that know the dynamical models exactly.

The Kalman filter (KF) and its variants are among the most celebrated algorithms in signal processing. These methods are used for state estimation of dynamic systems by relying on mathematical representations in the form of simple state-space (SS) models, which may be crude and inaccurate descriptions of the underlying dynamics. Emerging data-centric artificial intelligence (AI) techniques tackle these tasks using deep neural networks (DNNs), which are model agnostic. Recent developments illustrate the possibility of fusing DNNs with classic Kalman-type filtering, obtaining systems that learn to track in partially known dynamics. This article provides a tutorial-style overview of design approaches for incorporating AI in aiding KF-type algorithms. We review both generic and dedicated DNN architectures suitable for state estimation and provide a systematic presentation of techniques for fusing AI tools with KFs and for leveraging partial SS modeling and data, categorizing design approaches into task oriented and SS model oriented. The usefulness of each approach in preserving the individual strengths of model-based KFs and data-driven DNNs is investigated in a qualitative and quantitative study (whose code is publicly available), illustrating the gains of hybrid model-based/data-driven designs. We also discuss existing challenges and future research directions that arise from fusing AI and Kalman-type algorithms.

Motivation: Inference of gene regulatory network (GRN) is challenging due to the inherent sparsity of the GRN matrix and noisy expression data, often leading to a high possibility of false positive or negative predictions. To address this, it is essential to leverage the sparsity of the GRN matrix and develop a robust method capable of handling varying levels of noise in the data. Moreover, most existing GRN inference methods produce only fixed point estimates, which lack the flexibility and informativeness for comprehensive network analysis. In contrast, a Bayesian approach that yields closed-form posterior distributions allows probabilistic link selection, offering insights into the statistical confidence of each possible link. Consequently, it is important to engineer a Bayesian GRN inference method and rigorously execute a benchmark evaluation compared to state-of-the-art methods.

Results: We propose a method—Bayesian inference of GRN via Sparse Modelling (BiGSM). BiGSM effectively exploits the sparsity of the GRN matrix and infers the posterior distributions of GRN links from noisy expression data by using the maximum likelihood based learning. We thoroughly benchmarked BiGSM using biological and simulated datasets including GeneNetWeaver, GeneSPIDER, and GRNbenchmark. The benchmark test evaluates its accuracy and robustness across varying noise levels and data models. Using point-estimate based performance measures, BiGSM provides an overall best performance in comparison with several state-of-the-art methods including GENIE3, LASSO, LSCON, and Zscore. Additionally, BiGSM is the only method in the set of competing methods that provides posteriors for the GRN weights, helping to decipher confidence across predictions.

Availability and implementation: Code implemented via MATLAB and Python are available at Github: https://github.com/SachLab/BiGSM and archived at zenodo.

Chagas disease (ChD) remains a major health burden in Latin America, and scalable screening tools are required to pre-select high-risk patients for confirmatory testing. Electrocardiograms (ECGs) are widely available, but machine learning models trained on heterogeneous datasets are vulnerable to domain bias. As part of the ‘Detection of Chagas Disease from the ECG: The George B. Moody PhysioNet Challenge 2025’, this paper presents a domain-aware framework using a transformer encoder initialized with released pretrained weights, combined with a four-class reformulation of SaMi-Trop and PTB-XL cohorts and a domain-adversarial head to promote domain-invariant features. In five-fold cross-validation, the model achieved an average challenge score (Recall@5%) of 0.824 (best 0.852, with AUROC/AUPRC of 0.883/0.852). Predicted probabilities remained conservative, avoiding extreme confidence while maintaining strong ranking. Submitted under the team name FjordNet, this model achieved an official test score of 0.173, ranking 31st out of 41 teams. These results demonstrate that the transformer encoder structure with multi-class adversarial training improve local robustness, but domain-aware validation and dataset diversification are necessary for generalizable ChD ECG screening.

IoT-based low-cost gas sensors networks are important for environmental monitoring, but their regular calibrations are needed to achieve acceptable sensing performance. A critical step in network calibration is identifying when sensors within the network are sensing the same phenomenon, which is essential for accurate calibration. In this paper, we propose an adaptive similarity-search-based method for detecting these periods of similarity under the assumption of linear sensor drift. Our method leverages the relationships between neighboring sensors' measurements to enhance calibration accuracy, outperforming the commonly used Pearson correlation approach. We validate the effectiveness of our method through experiments with both synthetic data and real-world CO2 sensor networks, demonstrating improved calibration accuracy and reliability.

We introduce a model-free hidden sequence (MHS) estimation problem where the task is to estimate a long sequence of 'model-free' process hidden under additive Gaussian noise. To estimate the posterior of the hidden sequence from the noisy observation sequence, we have three main challenges: (a) the process is 'model-free', that means there is no underlying state-space-model (SSM), limiting the use of traditional SSM-informed model-based methods like Kalman filter (KF) and particle filter (PF); (b) only an unlabelled dataset comprised of noisy observation sequences is available as training data, and hence no supervised learning possible; and,

In this article, we provide an experimental observation: Deep neural network (DNN) based speech quality assessment (SQA) models have inherent latent representations where many types of impairments are clustered. While DNN-based SQA models are not trained for impairment classification, our experiments show good impairment classification results in an appropriate SQA latent representation. We investigate the clustering of impairments using various kinds of audio degradations that include different types of noises, waveform clipping, gain transition, pitch shift, compression, reverberation, etc. To visualize the clusters we perform classification of impairments in the SQA-latent representation domain using a standard k-nearest neighbor (kNN) classifier. We also develop a new DNN-based SQA model, named DNSMOS+, to examine whether an improvement in SQA leads to an improvement in impairment classification. The classification accuracy is 94% for LibriAugmented dataset with 16 types of impairments and 54% for ESC-50 dataset with 50 types of real noises.