Abstract
Counterfactual explanations (CFs) help users understand and act on black-box machine learning decisions by suggesting minimal changes to achieve a desired outcome. However, existing methods often ignore individual feasibility, leading to unrealistic or unactionable recommendations. We propose a personalized CF generation method based on cluster-specific fine-tuning of Generative Adversarial Networks (GANs). By grouping users with similar behavior and constraints, we adapt immutable features and cost weights per cluster, allowing GANs to generate more actionable and user-aligned counterfactuals. Experiments on the German Credit dataset show that our approach achieves a 6x improvement in prediction gain and a 30% reduction in sparsity compared to a baseline CounterGAN, while maintaining plausibility and acceptable latency for online use.

Abstract
Effective warehouse management plays a pivotal role in optimizing supply chain performance, particularly in high-demand, time-sensitive environments. This study introduces a simulation-based decision support system designed to improve the management of Picking-By-Line (PBL) operations in cross-docking distribution centres. Developed in FlexSim and calibrated with empirical data from an industrial case study, the model replicates real-world warehouse conditions and is validated against observed operational performance. The tool supports warehouse managers in evaluating and comparing operational strategies, such as dynamic storage allocation policies and picker routing constraints, with the goal of reducing operator travel distances, mitigating congestion, and enhancing overall efficiency. A key contribution of this work is the integration of congestion-sensitive performance indicators that allow for a detailed analysis of the trade-offs between travel efficiency and localized congestion-an aspect often overlooked in traditional optimization methods. This study demonstrates the value of simulation as a scalable and realistic decision-support tool for optimizing PBL operations in complex and variable environments where human movement is a major cost and performance driver. The proposed tool bridges the gap between theoretical modelling and practical implementation, offering actionable insights for warehouse layout, space utilization, and resource allocation.

Abstract
Haemoglobin (Hb) concentration is a key biomarker for several diseases. Traditional laboratory methods often have limitations due to their time-consuming nature, the need for skilled personnel, or the use of high-cost instrumentation. This work presents a sensing strategy for developing new point-of-care tests (POCTs) for Hb detection via a proof of concept. The proposed sensing approach is implemented using plasmonic plastic optical fiber (POF) sensor chips that integrate an electropolymerized molecularly imprinted polymer (eMIP) film on the plasmonic surface for Hb-selective detection. The developed sensor system demonstrates an ultra-low detection limit of 80 fM in buffer, about five orders of magnitude lower than that of other comparable Hb sensors. Selectivity tests against common interfering proteins, such as bovine serum albumin (BSA) and immunoglobulin G (IgG), confirmed high specificity towards the target analyte. Moreover, the sensor's performance was tested using a whole-blood sample, yielding results consistent with those of standard haematology analysis. The proposed sensor system, based on simple equipment, provides a quick (about 10 min) and cost-effective (about 10 euros per chip) label-free diagnostic tool for POCTs in real-world scenarios, such as finger-prick sampling, offering a less invasive alternative to traditional laboratory methods, towards devices useful for Internet of Medical Things (IoMT).

Abstract

Abstract
Highlights What are the main findings? PDAM improves hyperspectral image classification robustness under noisy labels. Prototype-guided dynamic masking and reconstruction outperform competing methods across multiple noise levels and datasets. What are the implications of the main findings? Prototype-based reliability estimation is effective for identifying trustworthy supervision in noisy hyperspectral data. Reliability-aware masking offers a practical strategy for robust remote-sensing learning with corrupted annotations.Highlights What are the main findings? PDAM improves hyperspectral image classification robustness under noisy labels. Prototype-guided dynamic masking and reconstruction outperform competing methods across multiple noise levels and datasets. What are the implications of the main findings? Prototype-based reliability estimation is effective for identifying trustworthy supervision in noisy hyperspectral data. Reliability-aware masking offers a practical strategy for robust remote-sensing learning with corrupted annotations.Abstract Existing noisy-label hyperspectral image classification (HSIC) methods usually address clean sample selection and representation regularization as separate problems, although the reliability of observed labels varies substantially across samples in hyperspectral data. This issue is amplified by mixed pixels, boundary ambiguity, spectral overlap, and limited labeled samples, which make hard clean samples difficult to distinguish from mislabeled ones. We therefore propose PDAM, a sample-reliability-guided training framework for noisy-label HSIC. The method first estimates feature-space class consistency by comparing each sample with the prototype of its observed class and converting this consistency into a reliability probability with a Gaussian mixture model. To reduce conservative false negatives, matched high-confidence selection is further used to recover hard but correctly labeled samples. The resulting reliability estimate then determines how strongly the observed label is trusted through target refinement and how strongly the input is perturbed through reliability-guided masking. Finally, masked reconstruction provides label-independent structural regularization so that uncertain samples can still contribute to spectral-spatial representation learning. Under the evaluated synthetic symmetric noise settings on the University of Pavia (UP), Salinas Valley (SV), and Kennedy Space Center (KSC) datasets, PDAM achieves the best OA and Kappa in most reported comparisons and improves robustness under both moderate and severe noise. At 30% noise, PDAM reaches 97.30% OA on UP, 98.13% OA on SV, and 95.37% OA on KSC. Ablation studies further support the necessity of reliability estimation, hard clean sample recovery, and reliability-guided supervision and regularization within this unified training mechanism.

Abstract