arXiv subjects

Yingdan Shi

Publications and source records attributed to Yingdan Shi.

7 recordsLinked to original sources

arXiv2026

On the Plasticity Collapse in Continual Machine Unlearning

Machine unlearning enables deep neural networks to selectively remove the influence of specific data in response to privacy and regulatory requirements. While prior work largely studies single-shot unlearning, real-world systems must accommodate continual unlearning, where multiple unlearning requests occur sequentially over time. In this work, we identify a fundamental limitation of this setting: plasticity collapse, a progressive breakdown in a model's ability to effectively forget. Through theoretical analysis of continual unlearning dynamics, we show that continual unlearning operations accumulate geometric constraints in parameter space, leading to saturated subspaces that restrict future updates. This structural effect induces two distinct failure modes: (1) Forward failure -- diminishing forgetting quality for subsequent tasks, and (2) Backward failure -- spontaneous re-memorization of previously forgotten information. Extensive experiments across multiple architectures, datasets, and methods in image classification confirm that plasticity collapse is not an artifact of specific implementations, but a pervasive phenomenon inherent to continual unlearning. Our findings reveal a critical barrier to the long-term reliability of machine unlearning systems and motivate the development of plasticity-preserving unlearning algorithms. Our code is available at https://github.com/TIML-Group/Continual-Machine-Unlearning-Plasticity-Collapse

cs.LG↗

arXiv2025

Exploring Nonlinear Pathway in Parameter Space for Machine Unlearning

Machine Unlearning (MU) aims to remove the information of specific training data from a trained model, ensuring compliance with privacy regulations and user requests. While one line of existing MU methods relies on linear parameter updates via task arithmetic, they suffer from weight entanglement. In this work, we propose a novel MU framework called Mode Connectivity Unlearning (MCU) that leverages mode connectivity to find an unlearning pathway in a nonlinear manner. To further enhance performance and efficiency, we introduce a parameter mask strategy that not only improves unlearning effectiveness but also reduces computational overhead. Moreover, we propose an adaptive adjustment strategy for our unlearning penalty coefficient to adaptively balance forgetting quality and predictive performance during training, eliminating the need for empirical hyperparameter tuning. Unlike traditional MU methods that identify only a single unlearning model, MCU uncovers a spectrum of unlearning models along the pathway. Overall, MCU serves as a plug-and-play framework that seamlessly integrates with any existing MU methods, consistently improving unlearning efficacy. Extensive experiments on the image classification task demonstrate that MCU achieves superior performance. The codes are available at https://github.com/TIML-Group/Mode-Connectivity-Unlearning.

cs.AI↗

arXiv2025

HODDI: A Dataset of High-Order Drug-Drug Interactions for Computational Pharmacovigilance

Drug-side effect research is vital for understanding adverse reactions arising in complex multi-drug therapies. However, the scarcity of higher-order datasets that capture the combinatorial effects of multiple drugs severely limits progress in this field. Existing resources such as TWOSIDES primarily focus on pairwise interactions. To fill this critical gap, we introduce HODDI, the first Higher-Order Drug-Drug Interaction Dataset, constructed from U.S. Food and Drug Administration (FDA) Adverse Event Reporting System (FAERS) records spanning the past decade, to advance computational pharmacovigilance. HODDI contains 109,744 records involving 2,506 unique drugs and 4,569 unique side effects, specifically curated to capture multi-drug interactions and their collective impact on adverse effects. Comprehensive statistical analyses demonstrate HODDI's extensive coverage and robust analytical metrics, making it a valuable resource for studying higher-order drug relationships. Evaluating HODDI with multiple models, we found that simple Multi-Layer Perceptron (MLP) can outperform graph models, while hypergraph models demonstrate superior performance in capturing complex multi-drug interactions, further validating HODDI's effectiveness. Our findings highlight the inherent value of higher-order information in drug-side effect prediction and position HODDI as a benchmark dataset for advancing research in pharmacovigilance, drug safety, and personalized medicine. The dataset and codes are available at https://github.com/TIML-Group/HODDI.

cs.LG↗

arXiv2025

Tackling Fake Forgetting through Uncertainty Quantification

Machine unlearning seeks to remove the influence of specified data from a trained model. While the unlearning accuracy provides a widely used metric for assessing unlearning performance, it falls short in assessing the reliability of forgetting. In this paper, we find that the forgetting data points misclassified by unlearning accuracy still have their ground truth labels included in the conformal prediction set from the uncertainty quantification perspective, leading to a phenomenon we term fake forgetting. To address this issue, we propose a novel metric CR, inspired by conformal prediction, that offers a more reliable assessment of forgetting quality. Building on these insights, we further propose an unlearning framework CPU that incorporates conformal prediction into the Carlini & Wagner adversarial attack loss, enabling the ground truth label to be effectively removed from the conformal prediction set. Through extensive experiments on image classification tasks, we demonstrate both the effectiveness of our proposed metric and the superior forgetting quality achieved by our framework. Code is available at https://github.com/TIML-Group/Conformal-Prediction-Unlearning.

cs.LG↗

arXiv2025

Watermarking Graph Neural Networks via Explanations for Ownership Protection

Graph Neural Networks (GNNs) are widely deployed in industry, making their intellectual property valuable. However, protecting GNNs from unauthorized use remains a challenge. Watermarking offers a solution by embedding ownership information into models. Existing watermarking methods have two limitations: First, they rarely focus on graph data or GNNs. Second, the de facto backdoor-based method relies on manipulating training data, which can introduce ownership ambiguity through misclassification and vulnerability to data poisoning attacks that can interrupt the backdoor mechanism. Our explanation-based watermarking inherits the strengths of backdoor-based methods (e.g., black-box verification) without data manipulation, eliminating ownership ambiguity and data dependencies. In particular, we watermark GNN explanations such that these explanations are statistically distinct from others, so ownership claims must be verified through statistical significance. We theoretically prove that, even with full knowledge of our method, locating the watermark is NP-hard. Empirically, our method demonstrates robustness to fine-tuning and pruning attacks. By addressing these challenges, our approach significantly advances GNN intellectual property protection.

cs.CR↗

arXiv2023

DySuse: Susceptibility Estimation in Dynamic Social Networks

Influence estimation aims to predict the total influence spread in social networks and has received surged attention in recent years. Most current studies focus on estimating the total number of influenced users in a social network, and neglect susceptibility estimation that aims to predict the probability of each user being influenced from the individual perspective. As a more fine-grained estimation task, susceptibility estimation is full of attractiveness and practical value. Based on the significance of susceptibility estimation and dynamic properties of social networks, we propose a task, called susceptibility estimation in dynamic social networks, which is even more realistic and valuable in real-world applications. Susceptibility estimation in dynamic networks has yet to be explored so far and is computationally intractable to naively adopt Monte Carlo simulation to obtain the results. To this end, we propose a novel end-to-end framework DySuse based on dynamic graph embedding technology. Specifically, we leverage a structural feature module to independently capture the structural information of influence diffusion on each single graph snapshot. Besides, {we propose the progressive mechanism according to the property of influence diffusion,} to couple the structural and temporal information during diffusion tightly. Moreover, a self-attention block {is designed to} further capture temporal dependency by flexibly weighting historical timestamps. Experimental results show that our framework is superior to the existing dynamic graph embedding models and has satisfactory prediction performance in multiple influence diffusion models.

cs.SI↗

arXiv2023

Fairness-aware Competitive Bidding Influence Maximization in Social Networks

Competitive Influence Maximization (CIM) has been studied for years due to its wide application in many domains. Most current studies primarily focus on the micro-level optimization by designing policies for one competitor to defeat its opponents. Furthermore, current studies ignore the fact that many influential nodes have their own starting prices, which may lead to inefficient budget allocation. In this paper, we propose a novel Competitive Bidding Influence Maximization (CBIM) problem, where the competitors allocate budgets to bid for the seeds attributed to the platform during multiple bidding rounds. To solve the CBIM problem, we propose a Fairness-aware Multi-agent Competitive Bidding Influence Maximization (FMCBIM) framework. In this framework, we present a Multi-agent Bidding Particle Environment (MBE) to model the competitors' interactions, and design a starting price adjustment mechanism to model the dynamic bidding environment. Moreover, we put forward a novel Multi-agent Competitive Bidding Influence Maximization (MCBIM) algorithm to optimize competitors' bidding policies. Extensive experiments on five datasets show that our work has good efficiency and effectiveness.

cs.SI↗