Machine unlearning aims to remove the information of specific data from a trained machine learning model while retaining its utility for the remaining data, so as to meet the requirements of privacy regulations. Existing unlearning methods often assume a balanced data distribution, but neglect the real-world, long-tailed scenarios, where the decision boundaries of tail classes are frequently distorted due to insufficient sample representation, thereby reducing the unlearning efficacy. In this paper, we propose the first Long-Tailed Machine Unlearning (LTMU) framework from a unified decision-boundary perspective. Our framework begins with a directional boundary repair scheme designed to enrich the distorted decision boundary of the tail class, and then develop a novel boundary retraction approach tailored for long-tailed unlearning, dispersing both the augmented and original features throughout the feature space. This bidirectional manipulation not only offers a unified interpretation of the relationship between long-tailed learning and unlearning, but also enables flexible control over both repair and unlearning processes through the generation of augmented features, thereby effectively accomplishing the long-tailed unlearning task. Extensive experiments across multiple datasets and neural network architectures demonstrate the effectiveness of our framework in achieving complete unlearning of tail classes in long-tailed distributions.
论文检索
输入标题、作者或关键词，从 7,537 篇学术成果中精准定位
Towards Interpretable Drug-Drug Interaction Prediction: A Graph-Based Approach with Molecular and Network-Level Explanations
PDF ↗Drug-drug interactions (DDIs) represent a critical challenge in pharmacology, often leading to adverse drug reactions with significant implications for patient safety and healthcare outcomes. While graph-based methods have achieved strong predictive performance, most approaches treat drug pairs independently, overlooking the complex, context-dependent interactions unique to drug pairs. Additionally, these models struggle to integrate biological interaction networks and molecular-level structures to provide meaningful mechanistic insights. In this study, we propose MolecBioNet, a novel graph-based framework that integrates molecular and biomedical knowledge for robust and interpretable DDI prediction. By modeling drug pairs as unified entities, MolecBioNet captures both macro-level biological interactions and micro-level molecular influences, offering a comprehensive perspective on DDIs. The framework extracts local subgraphs from biomedical knowledge graphs and constructs hierarchical interaction graphs from molecular representations, leveraging classical graph neural network methods to learn multi-scale representations of drug pairs. To enhance accuracy and interpretability, MolecBioNet introduces two domain-specific pooling strategies: context-aware subgraph pooling (CASPool), which emphasizes biologically relevant entities, and attention-guided influence pooling (AGIPool), which prioritizes influential molecular substructures. The framework further employs mutual information minimization regularization to enhance information diversity during embedding fusion. Experimental results demonstrate that MolecBioNet outperforms state-of-the-art methods in DDI prediction, while ablation studies and embedding visualizations further validate the advantages of unified drug pair modeling and multi-scale knowledge integration. Moreover, the pooling strategies yield interpretable insights into the underlying interaction mechanisms, highlighting MolecBioNet's potential as a robust and transparent framework for predictive modeling in support of drug safety assessment.
Large language models (LLMs) have demonstrated their strong capabilities in various domains, and have been recently integrated for graph analysis as graph language models (GLMs). With LLMs as the predictor, some GLMs can interpret unseen tasks described by natural language, and learn from a few examples in the prompts without parameter tuning, known as in-context learning (ICL). Another subset of GLMs utilizes abundant training labels to enhance model performance, known as instruction tuning. However, we argue that ICL on graphs has effectiveness issues due to fixed parameters and efficiency issues due to long context. Meanwhile, the large amount of labeled data required for instruction tuning can be difficult to obtain in real-world scenarios. To this end, we aim to introduce an extra parameter adaptation stage that can efficiently tailor GLMs to an unseen graph and task with only a few labeled examples, in exchange for better prediction accuracy and faster inference speed. For implementation, in this paper we propose GraphLAMA method, with its model backbone and learning schemes specialized for efficient tuning and inference. Specifically, for the model backbone, we use a graph neural network (GNN) with several well-designed components (e.g., hop encodings, gating modules) to transform nodes into the representation space of LLM tokens. Task instructions can then be represented as a mixture of node and language tokens. In the pre-training stage, all model parameters except for the LLM will be trained with different tasks (i.e., node matching, node classification, and link prediction) to capture general knowledge. In the adaptation stage, only a few pre-trained parameters will be updated based on few-shot examples. Extensive experiments on few/zero-shot node classification and summary generation show that our proposed GraphLAMA achieves state-of-the-art (SOTA) performance with 4.91% absolute improvement in accuracy. Compared with ICL, our inference speed can be 10 times faster under 5-shot setting. Our code is available on GitHub at https://github.com/BUPT-GAMMA/GraphLAMA.
Biomedical knowledge graphs (KGs) encode rich, structured information critical for drug discovery tasks, but extracting meaningful insights from large-scale KGs remains challenging due to their complex structure. Existing biomedical subgraph retrieval methods are tailored for graph neural networks (GNNs), limiting compatibility with other paradigms, including large language models (LLMs). We introduce K-Paths, a model-agnostic retrieval framework that extracts structured, diverse, and biologically meaningful multi-hop paths from dense biomedical KGs. These paths enable prediction of unobserved drug-drug and drug-disease interactions, including those involving entities not seen during training, thus supporting inductive reasoning. K-Paths is training-free and employs a diversity-aware adaptation of Yen's algorithm to extract the K shortest loopless paths between entities in a query, prioritizing biologically relevant and relationally diverse connections. These paths serve as concise, interpretable reasoning chains that can be directly integrated with LLMs or GNNs to improve generalization, accuracy, and enable explainable inference. Experiments on benchmark datasets show that K-Paths improves zero-shot reasoning across state-of-the-art LLMs. For instance, Tx-Gemma 27B improves by 19.8 and 4.0 F1 points on interaction severity prediction and drug repurposing tasks, respectively. Llama 70B achieves gains of 8.5 and 6.2 points on the same tasks. K-Paths also boosts the training efficiency of EmerGNN, a state-of-the-art GNN, by reducing the KG size by 90% while maintaining predictive performance. Beyond efficiency, K-Paths bridges the gap between KGs and LLMs, enabling scalable and explainable LLM-augmented scientific discovery. We release our code and the retrieved paths as a benchmark for inductive reasoning.
The prevalence of unhealthy eating habits has become a growing concern in the United States. However, popular food recommendation platforms, such as Yelp, tend to prioritize users' dietary preferences over the healthiness of their choices. While some efforts have focused on developing health-aware food recommendation systems, personalization based on specific health conditions remains underexplored. Additionally, the lack of interpretability in these systems prevents users from evaluating the reliability of recommendations, limiting their practical adoption. To address these issues, we introduce two large-scale personalized health-aware food recommendation benchmarks at the first attempt. Building on this, we propose a novel framework called the Multi-Objective Personalized Interpretable Health-aware Food Recommendation System (MOPI-HFRS). This system generates food recommendations by jointly optimizing three objectives: user preference, personalized healthiness, and nutritional diversity. It also incorporates a reasoning module enhanced by large language models (LLMs) to provide interpretable recommendations that promote healthy dietary knowledge. The framework integrates descriptive features and health data using two structure learning and pooling modules within a graph learning framework. Pareto optimization is applied to balance the multi-faceted objectives. To further enhance healthy dietary knowledge, the system leverages LLMs by infusing knowledge from the recommendation model, generating meaningful interpretations for the recommendations. Extensive experiments on the proposed benchmarks demonstrate that MOPI-HFRS outperforms state-of-the-art methods by delivering diverse, healthy food recommendations alongside reliable explanations.
Traffic Signal Control (TSC) is a crucial component in urban traffic management, aiming to optimize road network efficiency and reduce congestion. Traditional TSC methods, primarily based on transportation engineering and reinforcement learning (RL), often struggle with generalization abilities across varied traffic scenarios and lack interpretability. This paper presents LLMLight, a novel framework employing Large Language Models (LLMs) as decision-making agents for TSC. Specifically, the framework begins by instructing the LLM with a knowledgeable prompt detailing real-time traffic conditions. Leveraging the advanced generalization capabilities of LLMs, LLMLight engages a reasoning and decision-making process akin to human intuition for effective traffic control. Moreover, we build LightGPT, a specialized backbone LLM tailored for TSC tasks. By learning nuanced traffic patterns and control strategies, LightGPT enhances the LLMLight framework cost-effectively. Extensive experiments conducted on ten real-world and synthetic datasets, along with evaluations by fifteen human experts, demonstrate the exceptional effectiveness, generalization ability, and interpretability of LLMLight with LightGPT, outperforming nine baseline methods and ten advanced LLMs. Our project is available at https://github.com/usail-hkust/LLMTSCS.
This paper presents an analytical study of the oversmoothing issue in diffusion-based Graph Neural Networks (GNNs). Generalizing beyond extant approaches grounded in random walk analysis or particle systems, we approach this problem through operator semigroup theory. This theoretical framework allows us to rigorously prove that oversmoothing is intrinsically linked to the ergodicity of the diffusion operator. Relying on semigroup method, we can quantitatively analyze the dynamic of graph diffusion and give a specific mathematical form of the smoothing feature by ergodicity and invariant measure of operator, which improves previous works only show existence of oversmoothing. This finding further poses a general and mild ergodicity-breaking condition, encompassing the various specific solutions previously offered, thereby presenting a more universal and theoretically grounded approach to relieve oversmoothing in diffusion-based GNNs. Additionally, we offer a probabilistic interpretation of our theory, forging a link with prior works and broadening the theoretical horizon. Our experimental results reveal that this ergodicity-breaking term effectively mitigates oversmoothing measured by Dirichlet energy, and simultaneously enhances performance in node classification tasks.
The widespread application of pre-trained language models (PLMs) in natural language processing (NLP) has led to increasing concerns about their explainability. Selective rationalization is a self-explanatory framework that selects human-intelligible input subsets as rationales for predictions. Recent studies have shown that applying existing rationalization frameworks to PLMs will result in severe degeneration and failure problems, producing sub-optimal or meaningless rationales. Such failures severely damage trust in rationalization methods and constrain the application of rationalization techniques on PLMs. In this paper, we find that the homogeneity of tokens in the sentences produced by PLMs is the primary contributor to these problems. To address these challenges, we propose a method named Pre-trained Language Model's Rationalization (PLMR), which splits PLMs into a generator and a predictor to deal with NLP tasks while providing interpretable rationales. The generator in PLMR also alleviates homogeneity by pruning irrelevant tokens, while the predictor uses full-text information to standardize predictions. Experiments conducted on two widely used datasets across multiple PLMs demonstrate the effectiveness of the proposed method PLMR in addressing the challenge of applying selective rationalization to PLMs. Codes: https://github.com/ylb777/PLMR.
Shapley values (Shap) are a popular approach to explaining decisions of black-box models by revealing the importance of individual features. Shap explanations are easy to interpret, but as they do not incorporate feature interactions, they are also incomplete and potentially misleading. Interaction-aware methods such as nShap report the additive importance of all subsets up to n features. These explanations are complete, but in practice excessively large and difficult to interpret. In this paper, we combine the best of both worlds. We partition the features into significantly interacting groups, and use these to compose a succinct, interpretable explanation. To determine which partitioning out of super-exponentially many explains a model best, we derive a criterion that weighs the complexity of an explanation against its representativeness for the model's behavior. To be able to find the best partitioning, we show how to prune sub-optimal solutions using a statistical test. This not only improves runtime but also helps to avoid explaining spurious interactions. Experiments show that iShap represents underlying modeling more accurately than Shap and nShap, and a user study suggests that iShap is perceived as more interpretable and trustworthy.
Graph Transformers (GTs) have recently achieved significant success in the graph domain by effectively capturing both long-range dependencies and graph inductive biases. However, these methods face two primary challenges: (1) multi-view chaos, which results from coupling multi-view information (positional, structural, attribute), thereby impeding flexible usage and the interpretability of the propagation process. (2) local-global chaos, which arises from coupling local message passing with global attention, leading to issues of overfitting and over-globalizing. To address these challenges, we propose a high-level decoupled perspective of GTs, breaking them down into three components and two interaction levels: positional attention, structural attention, and attribute attention, alongside local and global interaction. Based on this decoupled perspective, we design a decoupled graph triple attention network named DeGTA, which separately computes multi-view attentions and adaptively integrates multi-view local and global information. This approach offers three key advantages: enhanced interpretability, flexible design, and adaptive integration of local and global information. Through extensive experiments, DeGTA achieves state-of-the-art performance across various datasets and tasks, including node classification and graph classification. Comprehensive ablation studies demonstrate that decoupling is essential for improving performance and enhancing interpretability. Our code is available at: https://github.com/wangxiaotang0906/DeGTA
Meta-learning is a practical learning paradigm to transfer skills across tasks from a few examples. Nevertheless, the existence of task distribution shifts tends to weaken meta-learners' generalization capability, particularly when the training task distribution is naively hand-crafted or based on simple priors that fail to cover critical scenarios sufficiently. Here, we consider explicitly generative modeling task distributions placed over task identifiers and propose robustifying fast adaptation from adversarial training. Our approach, which can be interpreted as a model of a Stackelberg game, not only uncovers the task structure during problem-solving from an explicit generative model but also theoretically increases the adaptation robustness in worst cases. This work has practical implications, particularly in dealing with task distribution shifts in meta-learning, and contributes to theoretical insights in the field. Our method demonstrates its robustness in the presence of task subpopulation shifts and improved performance over SOTA baselines in extensive experiments. The code is available at the project site https://sites.google.com/view/ar-metalearn.
Survival analysis is widely used as a technique to model time-to-event data when some data is censored, particularly in healthcare for predicting future patient risk. In such settings, survival models must be both accurate and interpretable so that users (such as doctors) can trust the model and understand model predictions. While most literature focuses on discrimination, interpretability is equally as important. A successful interpretable model should be able to describe how changing each feature impacts the outcome, and should only use a small number of features. In this paper, we present DyS (pronounced "dice''), a new survival analysis model that achieves both strong discrimination and interpretability. DyS is a feature-sparse Generalized Additive Model, combining feature selection and interpretable prediction into one model. While DyS works well for all survival analysis problems, it is particularly useful for large (in n and p) survival datasets such as those commonly found in observational healthcare studies. Empirical studies show that DyS competes with other state-of-the-art machine learning models for survival analysis, while being highly interpretable.
Counterfactual explanations (CFEs) guide users on how to adjust inputs to machine learning models to achieve desired outputs. While existing research primarily addresses static scenarios, real-world applications often involve data or model changes, potentially invalidating previously generated CFEs and rendering user-induced input changes ineffective. Current methods addressing this issue often support only specific models or change types, require extensive hyperparameter tuning, or fail to provide probabilistic guarantees on CFE robustness to model changes. This paper proposes a novel approach for generating CFEs that provides probabilistic guarantees for any model and change type, while offering interpretable and easy-to-select hyperparameters. We establish a theoretical framework for probabilistically defining robustness to model change and demonstrate how our BetaRCE method directly stems from it. BetaRCE is a post-hoc method applied alongside a chosen base CFE generation method to enhance the quality of the explanation beyond robustness. It facilitates a transition from the base explanation to a more robust one with user-adjusted probability bounds. Through experimental comparisons with baselines, we show that BetaRCE yields robust, most plausible, and closest to baseline counterfactual explanations.
Data-centric methods have shown great potential in understanding and predicting spatiotemporal dynamics, enabling better design and control of the object system. However, deep learning models often lack interpretability, fail to obey intrinsic physics, and struggle to cope with the various domains. While geometry-based methods, e.g., graph neural networks (GNNs), have been proposed to further tackle these challenges, they still need to find the implicit physical laws from large datasets and rely excessively on rich labeled data. In this paper, we herein introduce the conservation-informed GNN (CiGNN), an end-to-end explainable learning framework, to learn spatiotemporal dynamics based on limited training data. The network is designed to conform to the general conservation law via symmetry, where conservative and non-conservative information passes over a multiscale space enhanced by a latent temporal marching strategy. The efficacy of our model has been verified in various spatiotemporal systems based on synthetic and real-world datasets, showing superiority over baseline models. Results demonstrate that CiGNN exhibits remarkable accuracy and generalizability, and is readily applicable to learning for prediction of various spatiotemporal dynamics in a spatial domain with complex geometry.
Large quantities of social activity data, such as weekly web search volumes and the number of new infections with infectious diseases, reflect peoples' interests and activities. It is important to discover temporal patterns from such data and to forecast future activities accurately. However, modeling and forecasting social activity data streams is difficult because they are high-dimensional and composed of multiple time-varying dynamics such as trends, seasonality, and interest diffusion. In this paper, we propose D-Tracker, a method for continuously capturing time-varying temporal patterns within social activity tensor data streams and forecasting future activities. Our proposed method has the following properties: (a) Interpretable: it incorporates the partial differential equation into a tensor decomposition framework and captures time-varying temporal patterns such as trends, seasonality, and interest diffusion between locations in an interpretable manner; (b) Automatic: it has no hyperparameters and continuously models tensor data streams fully automatically; (c) Scalable: the computation time of D-Tracker is independent of the time series length. Experiments using web search volume data obtained from GoogleTrends, and COVID-19 infection data obtained from COVID-19 Open Data Repository show that our method can achieve higher forecasting accuracy in less computation time than existing methods while extracting the interest diffusion between locations. Our source code and datasets are available at https://github.com/Higashiguchi-Shingo/D-Tracker.
Cognitive diagnosis is a fundamental task in intelligent education, aiming to measure students' proficiency on knowledge concepts based on practice data. Traditional methods utilize a broadly-defined latent trait θ to represent knowledge proficiency with some cognitive factors like skill or ability. However, existing methods simplify this to a narrowly-defined latent trait θ, which focuses only on knowledge or treats these cognitive factors as implicit features inferred from data. They fail to explicitly model these cognitive factors, resulting in limited performance and interpretability. To this end, we revisit essence of cognition in Educational Psychology Theory and propose a novel Cognition-aware Cognitive Diagnosis (CCD) model, where we first introduce the Cognition factor as a bridge into the long-standing three-basic-factors (Student, Exercise, Knowledge concept) paradigm. CCD has two main parts: cognition representations and a two-stage diagnostic process. In the first part, we explicitly model cognitive process (CP) dimensions from Bloom's Taxonomy of Educational Objectives, leading to two innovative concepts proposed: the student's Subjective Cognitive Ability (SCA) and the exercise's Objective Cognitive Attribute (OCA), derived by regulating the CP through S-K and E-K interactions, respectively. Then, the SCA and OCA are formed into a new cognition-aware latent trait θ. In the second part, we employ a basic interaction function and a slip and guess influence function, inputting our new θ, a continuous Q-matrix (generated by a siamese PLMs), and other features to obtain the ideal result, followed by feeding it into the slip and guess influence function to obtain the actual result. Extensive experiments on real-world datasets demonstrates the superior effectiveness and good interpretability.
The ability to detect and adapt to changes in data distributions is crucial to maintain the accuracy and reliability of machine learning models. Detection is generally approached by observing the drift of model performance from a global point of view. However, drifts occurring in (fine-grained) data subgroups may go unnoticed when monitoring global drift.We take a different perspective, and introduce methods for observing drift at the finer granularity of subgroups. Relevant data subgroups are identified during training and monitored efficiently throughout the model's life. Performance drifts in any subgroup are detected, quantified and characterized so as to provide an interpretable summary of the model behavior over time. Experimental results confirm that our subgroup-level drift analysis identifies drifts that do not show at the (coarser) global dataset level. The proposed approach provides a valuable tool for monitoring model performance in dynamic real-world applications, offering insights into the evolving nature of data and ultimately contributing to more robust and adaptive models.
Road traffic forecasting is crucial in real-world intelligent transportation scenarios like traffic dispatching and path planning in city management and personal traveling. Spatio-temporal graph neural networks (STGNNs) stand out as the mainstream solution in this task. Nevertheless, the quadratic complexity of remarkable dynamic spatial modeling-based STGNNs has become the bottleneck over large-scale traffic data. From the spatial data management perspective, we present a novel Transformer framework called PatchSTG to efficiently and dynamically model spatial dependencies for large-scale traffic forecasting with interpretability and fidelity. Specifically, we design a novel irregular spatial patching to reduce the number of points involved in the dynamic calculation of Transformer. The irregular spatial patching first utilizes the leaf K-dimensional tree (KDTree) to recursively partition irregularly distributed traffic points into leaf nodes with a small capacity, and then merges leaf nodes belonging to the same subtree into occupancy-equaled and non-overlapped patches through padding and backtracking. Based on the patched data, depth and breadth attention are used interchangeably in the encoder to dynamically learn local and global spatial knowledge from points in a patch and points with the same index of patches. Experimental results on four real world large-scale traffic datasets show that our PatchSTG achieves train speed and memory utilization improvements up to 10x and 4x with the state-of-the-art performance.
The rise of cryptocurrencies like Bitcoin has not only increased trade volumes but also broadened the use of graph machine learning techniques, such as address embeddings, to analyze transactions and decipher user patterns. Traditional analysis methods rely on simple heuristics and extensive data gathering, while more advanced Graph Neural Networks encounter challenges such as scalability, poor interpretability, and label scarcity in massive blockchain transaction networks. To overcome existing techniques' computational and interpretability limitations, we introduce a topological approach, Chainlet Orbits, which embeds blockchain addresses by leveraging their topological characteristics in temporal transactions. We employ our innovative address embeddings to investigate financial behavior and e-crime in the Bitcoin and Ethereum networks, focusing on distinctive substructures that arise from user behavior. Our model demonstrates exceptional performance in node classification experiments compared to GNN-based approaches. Furthermore, our approach embeds all daily nodes of the largest blockchain transaction network, Bitcoin, and creates explainable machine learning models in less than 17 minutes which takes days for GNN-based approaches.
TRIKOP: Exploring Visual Prompting Paradigms for Multi-Grade Knee Osteoarthritis Classification on MRI Images
PDF ↗Knee osteoarthritis (KOA) is a degenerative joint disease that significantly impacts quality of life. While transfer learning shows promise in medical imaging, its application to KOA diagnosis remains challenging due to medical data's unique characteristics. To address this, we propose TRIKOP, a framework leveraging Visual Prompting for KOA diagnosis on MRI. Our approach explores three prompt-generating strategies that extract clinically relevant information from input images. Each prompt type is encoded using a tailored method to integrate effectively into the Vision Transformer for optimal representation. Among them, the contrastive embedding prompting strategy achieves 63.04% accuracy on the OAI dataset, surpassing prior studies. Moreover, TRIKOP produces attention maps highlighting diagnostically significant regions, improving model interpretability. This work highlights TRIKOP’s potential to improve AI-driven KOA diagnosis and clinical support.