However, SSA T suffers from catastrophic overfit-ting (CO), a phenomenon that leads to a severely distorted classifier, making it vulnerable to multi-step adversarial attacks. In this work, we observe that some adversarial examples generated on the SSA T -trained network exhibit anomalous behaviour, that is, although these training samples are generated by the inner maximization process, their associated loss decreases instead, which we named abnormal adversarial examples (AAEs).

This paper proposes a method for hiding the least-important samples during the training of deep neural networks to increase efficiency, i.e., to reduce the cost of

However, the leap in performance often comes at the cost of large-scale labeled data.

Recent advances in deep learning have improved multivariate time series (MTS) classification and regression by capturing complex patterns, but their lack of transparency hinders decision-making. Explainable AI (XAI) methods offer partial insights, yet often fall short of conveying the full decision space. Counterfactual Explanations (CE) provide a promising alternative, but current approaches typically prioritize either accuracy, proximity or sparsity -- rarely all -- limiting their practical value. To address this, we propose CONFETTI, a novel multi-objective CE method for MTS. CONFETTI identifies key MTS subsequences, locates a counterfactual target, and optimally modifies the time series to balance prediction confidence, proximity and sparsity. This method provides actionable insights with minimal changes, improving interpretability, and decision support. CONFETTI is evaluated on seven MTS datasets from the UEA archive, demonstrating its effectiveness in various domains. CONFETTI consistently outperforms state-of-the-art CE methods in its optimization objectives, and in six other metrics from the literature, achieving $\geq10\%$ higher confidence while improving sparsity in $\geq40\%$.

However, SSA T suffers from catastrophic overfit-ting (CO), a phenomenon that leads to a severely distorted classifier, making it vulnerable to multi-step adversarial attacks. In this work, we observe that some adversarial examples generated on the SSA T -trained network exhibit anomalous behaviour, that is, although these training samples are generated by the inner maximization process, their associated loss decreases instead, which we named abnormal adversarial examples (AAEs).

This paper proposes a method for hiding the least-important samples during the training of deep neural networks to increase efficiency, i.e., to reduce the cost of

We thank all reviewers for their constructive feedback! We address reviewers comments below, and will incorporate all feedback. This explains why SD outperforms LS. Please refer to our response to [R3] for discussion on CD. One can alternatively compute the variance of prediction confidence.

We make three contributions to conformal prediction. First, we propose fuzzy conformal confidence sets that offer a degree of exclusion, generalizing beyond the binary inclusion/exclusion offered by classical confidence sets. We connect fuzzy confidence sets to e-values to show this degree of exclusion is equivalent to an exclusion at different confidence levels, capturing precisely what e-values bring to conformal prediction. We show that a fuzzy confidence set is a predictive distribution with a more appropriate error guarantee. Second, we derive optimal conformal confidence sets by interpreting the minimization of the expected measure of the confidence set as an optimal testing problem against a particular alternative. We use this to characterize exactly in what sense traditional conformal prediction is optimal. Third, we generalize the inheritance of guarantees by subsequent minimax decisions from confidence sets to fuzzy confidence sets. All our results generalize beyond the exchangeable conformal setting to prediction sets for arbitrary models. In particular, we find that any valid test (e-value) for a hypothesis automatically defines a (fuzzy) prediction confidence set.