Clustering non-independent and identically distributed (non-IID) data under local differential privacy (LDP) in federated settings presents a critical challenge: preserving privacy while maintaining accuracy without iterative communication. Existing one-shot methods rely on unstable pairwise centroid distances or neighborhood rankings, degrading severely under strong LDP noise and data heterogeneity. We present Gravitational Federated Clustering (GFC), a novel approach to privacy-preserving federated clustering that overcomes the limitations of distance-based methods under varying LDP. Addressing the critical challenge of clustering non-IID data with diverse privacy guarantees, GFC transforms privatized client centroids into a global gravitational potential field where true cluster centers emerge as topologically persistent singularities. Our framework introduces two key innovations: (1) a client-side compactness-aware perturbation mechanism that encodes local cluster geometry as "mass" values, and (2) a server-side topological aggregation phase that extracts stable centroids through persistent homology analysis of the potential field's superlevel sets. Theoretically, we establish a closed-form bound between the privacy budget ε and centroid estimation error, proving the potential field's Lipschitz smoothing properties exponentially suppress noise in high-density regions. Empirically, GFC outperforms state-of-the-art methods on ten benchmarks, especially under strong LDP constraints (ε < 1), while maintaining comparable performance at lower privacy budgets. By reformulating federated clustering as a topological persistence problem in a synthetic physics-inspired space, GFC achieves unprecedented privacy-accuracy trade-offs without iterative communication, providing a new perspective for privacy-preserving distributed learning.

Link prediction plays a vital role in uncovering hidden relationships within complex networks, enabling applications such as identifying potential customers and products. However, this task faces critical challenges, including growing concerns over data privacy and the substantial computational and storage costs associated with large-scale networks. Condensed graphs, which are significantly smaller yet retain essential structural information, have emerged as a promising solution for preserving data utility while enhancing privacy. Despite this potential, existing methods like HyDRO rely on random node selection strategies designed primarily for node classification, overlooking connectivity patterns crucial for effective link prediction. Moreover, these methods lack rigorous evaluations of privacy risks associated with nodes and links in the condensed graphs—an essential consideration for sensitive real-world networks, where privacy concerns far exceed those of public datasets such as citation graphs. To address these limitations, we introduce HyDRO + , a novel graph condensation method guided by algebraic Jaccard similarity. By leveraging local connectivity patterns, HyDRO + generates structurally-aware condensed graphs that preserve link information more effectively. We further introduce a comprehensive evaluation framework that rigorously assesses both node- and link-level privacy leakage. Extensive experiments on four real-world networks demonstrate that HyDRO + consistently outperforms state-of-the-art methods—and even the original networks—in balancing link prediction accuracy and privacy preservation. Notably, it achieves nearly 20 × faster training and reduces storage requirements by a factor of 452 on the Computers dataset. This work represents the first attempt to use condensed graphs for privacy-preserving link prediction in real-world complex networks, offering a practical and scalable solution for secure information sharing in large-scale networks.

Machine unlearning, the ability for a machine learning model to forget, is becoming increasingly important to comply with data privacy regulations, as well as to remove harmful, manipulated, or outdated information. The key challenge lies in forgetting specific information while protecting model performance on the remaining data. While current state-of-the-art methods perform well, they typically require some level of retraining over the retained data, in order to protect or restore model performance. This adds computational overhead and mandates that the training data remain available and accessible, which may not be feasible. In contrast, other methods employ a retrain-free paradigm, however, these approaches are prohibitively computationally expensive and do not perform on par with their retrain-based counterparts. We present Selective Synaptic Dampening (SSD), a novel two-step, post hoc, retrain-free approach to machine unlearning which is fast, performant, and does not require long-term storage of the training data. First, SSD uses the Fisher information matrix of the training and forgetting data to select parameters that are disproportionately important to the forget set. Second, SSD induces forgetting by dampening these parameters proportional to their relative importance to the forget set with respect to the wider training data. We evaluate our method against several existing unlearning methods in a range of experiments using ResNet18 and Vision Transformer. Results show that the performance of SSD is competitive with retrain-based post hoc methods, demonstrating the viability of retrain-free post hoc unlearning approaches.

Organizations often struggle to identify the causes of change in metrics, such as product quality and delivery duration. This task becomes increasingly challenging when the cause lies outside of company borders in multiechelon supply chains that are only partially observable. Although traditional supply chain management has advocated for data sharing to gain better insights, this does not take place in practice due to data privacy concerns. We propose the use of explainable artificial intelligence for decentralized computing of estimated contributions to a metric of interest in a multistage production process. This approach mitigates the need to convince supply chain actors to share data, as all computations occur in a decentralized manner. Our method is empirically validated using data collected from a real multistage manufacturing process. The results demonstrate the effectiveness of our approach in detecting the source of quality variations compared to a centralized approach using Shapley additive explanations.

Despite numerous studies of deep autoencoders (AEs) for unsupervised anomaly detection, AEs still lack a way to express uncertainty in their predictions, crucial for ensuring safe and trustworthy machine learning systems in high-stake applications. Therefore, in this work, the formulation of Bayesian autoencoders (BAEs) is adopted to quantify the total anomaly uncertainty, comprising epistemic and aleatoric uncertainties. To evaluate the quality of uncertainty, we consider the task of classifying anomalies with the additional option of rejecting predictions of high uncertainty. In addition, we use the accuracy-rejection curve and propose the weighted average accuracy as a performance metric. Our experiments demonstrate the effectiveness of the BAE and total anomaly uncertainty on a set of benchmark datasets and two real datasets for manufacturing: one for condition monitoring, the other for quality inspection.

Understanding the uncertainty of a neural network’s (NN) predictions is essential for many purposes. The Bayesian framework provides a principled approach to this, however applying it to NNs is challenging due to large numbers of parameters and data. Ensembling NNs provides an easily implementable, scalable method for uncertainty quantification, however, it has been criticised for not being Bayesian. This work proposes one modification to the usual process that we argue does result in approximate Bayesian inference; regularising parameters about values drawn from a distribution which can be set equal to the prior. A theoretical analysis of the procedure in a simplified setting suggests the recovered posterior is centred correctly but tends to have an underestimated marginal variance, and overestimated correlation. However, two conditions can lead to exact recovery. We argue that these conditions are partially present in NNs. Empirical evaluations demonstrate it has an advantage over standard ensembling, and is competitive with variational methods.

This paper considers the generation of prediction intervals (PIs) by neural networks for quantifying uncertainty in regression tasks. It is axiomatic that high-quality PIs should be as narrow as possible, whilst capturing a specified portion of data. We derive a loss function directly from this axiom that requires no distributional assumption. We show how its form derives from a likelihood principle, that it can be used with gradient descent, and that model uncertainty is accounted for in ensembled form. Benchmark experiments show the method outperforms current state-of-the-art uncertainty quantification methods, reducing average PI width by over 10%.

Self-serving assets (SSAs) are a new interpretation of the intelligent product technology, set to transform product lifecycle management through automation. SSAs are engineering assets that autonomously monitor their health and expiry dates, search for suppliers, and negotiate with them, while they are still in use by the customer. The concept enables more timely and transparent supplier decision making while eliminating central database transactions and tedious manual effort. Autonomous self-interested agents that act on behalf of their stakeholders naturally give rise to an allocation problem, under the assumption of private information held by trade parties and capacity constrained suppliers providing imperfectly substitutable goods (ISGs). In this paper, we develop and compare three automated competition mechanisms, constructed as iterative games, and test them in the context of the aerospace service supply chain. The competition mechanisms include a prioritized selection mechanism, extended Vickrey, and reverse Dutch auctions. Our context drives us to seek mechanisms that will not only perform well in terms of economic theory, but also in terms of computational performance. Key findings are that extended Vickrey auctions can handle multiple criteria and provide higher market efficiency at lower computational cost, especially in small to medium markets. As scalability is an issue in large markets, the use of auctions is recommended only for complex high value assets or under uncertain market scenarios. As business-to-business (B2B) environments are becoming the norm for many global companies, our study aims to be exemplary to those who would like to implement automated auction mechanisms in highly complex environments.

This paper emphasizes the necessity of formally bringing qualitative and quantitative criteria of ergonomic design together, and provides a novel complementary design framework with this aim. Within this framework, different design criteria are viewed as optimization objectives, and design solutions are iteratively improved through the cooperative efforts of computer and user. The framework is rooted in multiobjective optimization, genetic algorithms, and interactive user evaluation. Three different algorithms based on the framework are developed, and tested with an ergonomic chair design problem. The parallel and multiobjective approaches show promising results in fitness convergence, design diversity, and user satisfaction metrics.

The concept of qualitativeness in design is an important one, and needs to be incorporated in the optimization process for a number of reasons outlined in this paper. interactive evolutionary computation and fuzzy systems are two of the widely used approaches for handling quanlitativeness in design optimization. This paper classifies the types of quantitativeness observed in design optimization, makes the case for their necessity, and proposes a novel framework for handling them, combining the two approaches in an evolutionary multi-objective optimization platform. Two components of the framework are tested using the floor-planning problem, and observations are reported. Future work is defined on the development of the framework.