论文检索

Chengkun Sun, Jinqian Pan, Russell Stevens Terry, Jiang Bian, Jie Xu

Generative models can enhance discriminative classifiers by constructing complex feature spaces, thereby improving performance on intricate datasets. Conventional methods typically augment datasets with more detailed feature representations or increase dimensionality to make nonlinear data linearly separable. Utilizing a generative model solely for feature space processing falls short of unlocking its full potential within a classifier and typically lacks a solid theoretical foundation. We base our approach on a novel hypothesis: the probability information (logit) derived from a single model training can be used to generate the equivalent of multiple training sessions. Leveraging the central limit theorem, this synthesized probability information is anticipated to converge toward the true probability more accurately. To achieve this goal, we propose the Bernoulli-Gaussian Decision Block (BGDB), a novel module inspired by the Central Limit Theorem and the concept that the mean of multiple Bernoulli trials approximates the probability of success in a single trial. Specifically, we utilize Improved Denoising Diffusion Probabilistic Models (IDDPM) to model the probability of Bernoulli Trials. Our approach shifts the focus from reconstructing features to reconstructing logits, transforming the logit from a single iteration into logits analogous to those from multiple experiments. We provide the theoretical foundations of our approach through mathematical analysis and validate its effectiveness through experimental evaluation using various datasets for multiple imaging tasks, including both classification and segmentation.

Chengkun Sun, Jinqian Pan, Zhuoli Jin, Russell Stevens Terry, Jiang Bian, Jie Xu

Training deep Convolutional Neural Networks (CNNs) presents unique challenges, including the pervasive issue of elimination singularities—consistent deactivation of nodes leading to degenerate manifolds within the loss landscape. These singularities impede efficient learning by disrupting feature propagation. To mitigate this, we introduce Pool Skip, an architectural enhancement that strategically combines a Max Pooling, a Max Unpooling, a 3 × 3 convolution, and a skip connection. This configuration helps stabilize the training process and maintain feature integrity across layers. We also propose the Weight Inertia hypothesis, which underpins the development of Pool Skip, providing theoretical insights into mitigating degradation caused by elimination singularities through dimensional and affine compensation. We evaluate our method on a variety of benchmarks, focusing on both 2D natural and 3D medical imaging applications, including tasks such as classification and segmentation. Our findings highlight Pool Skip's effectiveness in facilitating more robust CNN training and improving model performance.

Bohang Sun, Yongjian Deng, Yuena Lin, Qiuru Hai, Zhen Yang, Gengyu Lyu

Federated Multi-View Clustering (FMVC) aims to learn a global clustering model from heterogeneous data distributed across different devices, where each device only stores one view of all clustering samples. The key to deal with such problem lies in how to effectively fuse these heterogeneous samples while strictly preserve the data privacy across multiple devices. In this paper, we propose a novel structural graph learning framework named MGCD, which leverages both consistency and diversity of multi-view graph structure across global view-fusion server and local view-specific clients to achieve desired clustering while better preserves data privacy. Specifically, in each local client, we design a dual autoencoder to extract the latent consensuses and specificities of each view, where self-representation construction is introduced to generate the corresponding view-specific diversity graph. In the global server, the consistency implied in uploaded diversity graphs are further distilled and then incorporated into the consistency graph for subsequent cross-view contrastive fusion. During the training process, the server generates a global consistency graph and distributes it to each client for assisting in diversity graph construction, while the clients extract view-specific information and upload it to the server for more reliable consistency graph generation. The ``server-client'' interaction is conducted in an iterative manner, where the consistency implied in each local client is gradually aggregated into the global consistency graph, and the final clustering results are obtained by spectral clustering on the desired global consistency graph. Extensive experiments on various datasets have demonstrated the effectiveness of our proposed method on clustering federated multi-view data.

Bernardo Subercaseaux, Marcelo Arenas, Kuldeep S. Meel

Formal XAI is an emerging field that focuses on providing explanations with mathematical guarantees for the decisions made by machine learning models. A significant amount of work in this area is centered on the computation of ``sufficient reasons''. Given a model M and an input instance x, a sufficient reason for the decision x is a subset S of the features of x such that for any instance z that has the same values as x for every feature in S, it holds that M(x) = M(z). Intuitively, this means that the features in S are sufficient to fully justify the classification of x by M. For sufficient reasons to be useful in practice, they should be as small as possible, and a natural way to reduce the size of sufficient reasons is to consider a probabilistic relaxation; the probability of M(x) = M(z) must be at least some value delta in (0,1], where z is a random instance that coincides with x on the features in S. Computing small delta-sufficient reasons (delta-SRs) is known to be a theoretically hard problem; even over decision trees — traditionally deemed simple and interpretable models — strong inapproximability results make the efficient computation of small delta-SRs unlikely. We propose the notion of (delta, epsilon)-SR, a simple relaxation of delta-SRs, and show that this kind of explanations can be computed efficiently over linear models.

Xihong Su, Marek Petrik, Julien Grand-Clément

Optimizing risk-averse objectives in discounted MDPs is challenging because most models do not admit direct dynamic programming equations and require complex history-dependent policies. In this paper, we show that the risk-averse total reward criterion, under the Entropic Risk Measure (ERM) and Entropic Value at Risk (EVaR) risk measures, can be optimized by a stationary policy, making it simple to analyze, interpret, and deploy. We propose exponential value iteration, policy iteration, and linear programming to compute optimal policies. Compared with prior work, our results only require the relatively mild condition of transient MDPs and allow for both positive and negative rewards. Our results indicate that the total reward criterion may be preferable to the discounted criterion in a broad range of risk-averse reinforcement learning domains.

Peng Su, Shudong Huang, Weihong Ma, Deng Xiong, Jiancheng Lv

Previous multi-view contrastive learning methods typically operate at two scales: instance-level and cluster-level. The former generally constructs positive and negative pairs based on the correspondence between samples and view instances. These methods aim to bring positive pairs closer and push negative pairs further apart in the latent space. This kind of approaches has the drawback of inevitably introducing false negatives in an unsupervised setting, leading to reduced model discriminability. The latter usually involves calculating cluster assignments for samples under each view and maximizing view consensus by reducing distribution discrepancies through methods like optimizing the KL divergence between different view distributions or maximizing mutual information. However, clusters represent a macro structure that overlooks the local structure within the sample set, and the relationships between clusters across different views cannot be explicitly measured. To overcome the shortcomings of these two types of methods, we propose a method named Multi-view Granular-ball Contrastive Clustering (MGBCC). This method segments the sample set into coarse-grained granular balls, and establishes associations between intra-view and cross-view granular balls. These associations are reinforced in a shared latent space, thereby achieving multi-granularity contrastive learning. Granular balls lie between instances and clusters, naturally preserving the local topological structure of the sample set. We conduct extensive experiments to validate the effectiveness of the proposed method.

Junwei Su, Shan Wu

Temporal Graph Neural Networks (TGNNs) are a family of graph neural networks designed to model and learn dynamic information from temporal graphs. Given their substantial empirical success, there is an escalating interest in TGNNs within the research community. However, the majority of these efforts have been channelled towards algorithm and system design, with the evaluation metrics receiving comparatively less attention. Effective evaluation metrics are crucial for providing detailed performance insights, particularly in the temporal domain. This paper investigates the commonly used evaluation metrics for TGNNs and illustrates the failure mechanisms of these metrics in capturing essential temporal structures in the predictive behaviour of TGNNs. We provide a mathematical formulation of existing performance metrics and utilize an instance-based study to underscore their inadequacies in identifying volatility clustering (the occurrence of emerging errors within a brief interval). This phenomenon has profound implications for both algorithm and system design in the temporal domain. To address this deficiency, we introduce a new volatility-aware evaluation metric (termed volatility cluster statistics), designed for a more refined analysis of model temporal performance. Additionally, we demonstrate how this metric can serve as a temporal-volatility-aware training objective to alleviate the clustering of temporal errors. Through comprehensive experiments on various TGNN models, we validate our analysis and the proposed approach. The empirical results offer revealing insights: 1) existing TGNNs are prone to making errors with volatility clustering, and 2) TGNNs with different mechanisms to capture temporal information exhibit distinct volatility clustering patterns. Moreover, our empirical findings demonstrate that our proposed training objective effectively reduces volatility clusters in error.

Huai-An Su, Jiaxiang Geng, Liang Li, Xiaoqi Qin, Yanzhao Hou, Hao Wang, Xin Fu, Miao Pan

As a popular distributed learning paradigm, federated learning (FL) over mobile devices fosters numerous applications, while their practical deployment is hindered by participating devices' computing and communication heterogeneity. Some pioneering research efforts proposed to extract subnetworks from the global model, and assign as large a subnetwork as possible to the device for local training based on its full computing capacity. Although such fixed size subnetwork assignment enables FL training over heterogeneous mobile devices, it is unaware of (i) the dynamic changes of devices' communication and computing conditions and (ii) FL training progress and its dynamic requirements of local training contributions, both of which may cause very long FL training delay. Motivated by those dynamics, in this paper, we develop a wireless and heterogeneity aware latency efficient FL (WHALE-FL) approach to accelerate FL training through adaptive subnetwork scheduling. Instead of sticking to the fixed size subnetwork, WHALE-FL introduces a novel subnetwork selection utility function to capture device and FL training dynamics, and guides the mobile device to adaptively select the subnetwork size for local training based on (a) its computing and communication capacity, (b) its dynamic computing and/or communication conditions, and (c) FL training status and its corresponding requirements for local training contributions. Our evaluation shows that, compared with peer designs, WHALE-FL effectively accelerates FL training without sacrificing learning accuracy.

Chao Su, Huiming Zheng, Dezhong Peng, Xu Wang

Cross-modal retrieval aims to retrieve relevant data across different modalities. Driven by costly massive labeled data, existing cross-modal retrieval methods achieve encouraging results. To reduce annotation costs while maintaining performance, this paper focuses on an untouched but challenging problem, i.e., cross-modal retrieval with partial labels (PLCMR). PLCMR faces the dual challenges of annotation ambiguity and modality gap. To address these challenges, we propose a novel method termed disambiguated contrastive alignment (DiCA) for cross-modal retrieval with partial labels. Specifically, DiCA proposes a novel non-candidate boosted disambiguation learning mechanism (NBDL), which elaborately balances the trade-off between the losses on candidate and non-candidate labels that eliminate label ambiguity and narrow the modality gap. Moreover, DiCA presents an instance-prototype representation learning mechanism (IPRL) to enhance the model by further eliminating the modality gap at both the instance and prototype levels. Thanks to NBDL and IPRL, our DiCA effectively addresses the issues of annotation ambiguity and modality gap for cross-modal retrieval with partial labels. Experiments on four benchmarks validate the effectiveness of our proposed method, which demonstrates enhanced performance over existing state-of-the-art methods.

Artyom Stitsyuk, Jaesik Choi

In recent years, the application of transformer-based models in time-series forecasting has received significant attention. While often demonstrating promising results, the transformer architecture encounters challenges in fully exploiting the temporal relations within time series data due to its attention mechanism. In this work, we design eXponential Patch (xPatch for short), a novel dual-stream architecture that utilizes exponential decomposition. Inspired by the classical exponential smoothing approaches, xPatch introduces the innovative seasonal-trend exponential decomposition module. Additionally, we propose a dual-flow architecture that consists of an MLP-based linear stream and a CNN-based non-linear stream. This model investigates the benefits of employing patching and channel-independence techniques within a non-transformer model. Finally, we develop a robust arctangent loss function and a sigmoid learning rate adjustment scheme, which prevent overfitting and boost forecasting performance.

Uri Stern, Tomer Yaacoby, Daphna Weinshall

The infrequent occurrence of overfitting in deep neural networks is perplexing: contrary to theoretical expectations, increasing model size often enhances performance in practice. But what if overfitting does occur, though restricted to specific sub-regions of the data space? In this work, we propose a novel score that captures the forgetting rate of deep models on validation data. We posit that this score quantifies local overfitting: a decline in performance confined to certain regions of the data space. We then show empirically that local overfitting occurs regardless of the presence of traditional overfitting. Using the framework of deep over-parametrized linear models, we offer a certain theoretical characterization of forgotten knowledge, and show that it correlates with knowledge forgotten by real deep models. Finally, we devise a new ensemble method that aims to recover forgotten knowledge, relying solely on the training history of a single network. When combined with knowledge distillation, this method will enhance the performance of a trained model without adding inference costs. Extensive empirical evaluations demonstrate the efficacy of our method across multiple datasets, contemporary neural network architectures, and training protocols.

Luca Pascal Staus, Christian Komusiewicz, Frank Sommer, Manuel Sorge

Decision trees are a classic model for summarizing and classifying data. To enhance interpretability and generalization properties, it has been proposed to favor small decision trees. Accordingly, in the minimum-size decision tree training problem (MSDT), the input is a set of training examples in $\mathbb{R}^d$ with class labels and we aim to find a decision tree that classifies all training examples correctly and has a minimum number of nodes. MSDT is NP-hard and therefore presumably not solvable in polynomial time. Nevertheless, a promising algorithmic paradigm called witness trees which solves MSDT efficiently if the solution tree is small has been developed. In this work, we test this paradigm empirically. We provide an implementation, augment it with extensive heuristic improvements, and scrutinize it on standard benchmark instances. The augmentations achieve a mean 324-fold (median 84-fold) speedup over the naive implementation. Compared to the state of the art they achieve a mean 32-fold (median 7-fold) speedup over the dynamic programming based MurTree solver and a mean 61-fold (median 25-fold) speedup over SAT-based implementations. As a theoretical result we obtain an improved worst-case running-time bound for MSDT.

Padmanaba Srinivasan, William Knottenbelt

Offline reinforcement learning (RL) methods aim to learn optimal policies with access only to trajectories in a fixed dataset. Policy constraint methods formulate policy learning as an optimization problem that balances maximizing reward with minimizing deviation from the behavior policy. Closed form solutions to this problem can be derived as weighted behavioral cloning objectives that, in theory, must compute an intractable partition function. Reinforcement learning has gained popularity in language modeling to align models with human preferences; some recent works consider paired completions that are ranked by a preference model following which the likelihood of the preferred completion is directly increased. We adapt this approach of paired comparison. By reformulating the paired-sample optimization problem, we fit the maximum-mode of the Q function while maximizing behavioral consistency of policy actions. This yields our algorithm, Behavior Preference Regression for offline RL (BPR). We empirically evaluate BPR on the widely used D4RL Locomotion and Antmaze datasets, as well as the more challenging V-D4RL suite, which operates in image-based state spaces. BPR demonstrates state-of-the-art performance over all domains. Our on-policy experiments suggest that BPR takes advantage of the stability of on-policy value functions with minimal performance degradation on Locomotion datasets.

Fabian Christian Spaeh, Alina Ene, Huy Nguyen

Constrained k-submodular maximization is a general framework that captures many discrete optimization problems such as ad allocation, influence maximization, personalized recommendation, and many others. In many of these applications, datasets are large or decisions need to be made in an online manner, which motivates the development of efficient streaming and online algorithms. In this work, we develop single-pass streaming and online algorithms for constrained k-submodular maximization with both monotone and general (possibly non-monotone) objectives subject to cardinality and knapsack constraints. Our algorithms achieve provable constant-factor approximation guarantees which improve upon the state of the art in almost all settings. Moreover, they achieve the fastest known running times and have optimal space usage. We experimentally evaluate our algorithms on instances for ad allocation and other applications, where we observe that our algorithms are practical and scalable, and construct solutions that are comparable in value even to offline greedy algorithms.

Tananun Songdechakraiwut, Yutong Wu

The human brain is a complex system, and understanding its mechanisms has been a long-standing challenge in neuroscience. The study of the functional connectome, which maps the functional connections between different brain regions, has provided valuable insights through various advanced analysis techniques developed over the years. Similarly, neural networks, inspired by the brain's architecture, have achieved notable success in diverse applications but are often noted for their lack of interpretability. In this paper, we propose a novel approach that bridges neural networks and human brain functions by leveraging brain-inspired techniques. Our approach, grounded in the insights from the functional connectome, offers scalable ways to characterize topology of large neural networks using stable statistical and machine learning techniques. Our empirical analysis demonstrates its capability to enhance the interpretability of neural networks, providing a deeper understanding of their underlying mechanisms.

Yeji Song, Jimyeong Kim, Wonhark Park, Wonsik Shin, Wonjong Rhee, Nojun Kwak

In a surge of text-to-image (T2I) models and their customization methods that generate new images of a user-provided subject, current works focus on alleviating the costs incurred by a lengthy per-subject optimization. These zero-shot customization methods encode the image of a specified subject into a visual embedding which is then utilized alongside the textual embedding for diffusion guidance. The visual embedding incorporates intrinsic information about the subject, while the textual embedding provides a new context. However, the existing methods often 1) generate images with the same pose as an input image, and 2) exhibit deterioration in the subject's identity when facing a pose variation prompt. We first pin down the problem and show that redundant pose information in the visual embedding interferes with the pose indication in the textual embedding. Conversely, the textual embedding also harms the subject's identity which is tightly entangled with the pose in the visual embedding. As a remedy, we propose text-orthogonal visual embedding which effectively harmonizes with the given textual embedding. We also adopt the visual-only embedding and inject the subject's clear features utilizing a self-attention swap. Our method is both effective and robust, offering highly flexible zero-shot generation while effectively maintaining the subject's identity.

Lei Song, Jiaxing Li, Qinghua Si, Shihan Guan, Youyong Kong

Catastrophic forgetting poses a significant challenge for graph neural networks in continuously updating their knowledge base with data streams. To address this issue, much of the research has focused on node-level continual learning using parameter regularization or rehearsal-based strategies, while little attention given to graph-level tasks. Furthermore, current paradigms for continual graph learning may inadvertently capture spurious correlations for specific tasks through shortcuts, thereby exacerbating the forgetting of previous knowledge when new tasks are introduced. To tackle these challenges, we propose a novel paradigm, Rationale Learning GNN (RL-GNN), for graph-level continual graph learning. Specifically, we harness the invariant learning principle to incorporate environmental interventions into both the current and historical distributions, aiming to uncover rationales by minimizing empirical risk across all environments. The rationale serves as the sole factor guiding the learning process. Therefore, continual graph learning is redefined as capturing these invariant rationales within task sequences, alleviating catastrophic forgetting caused by spurious features. Extensive experiments on real-world datasets with varying task lengths demonstrate the effectiveness of our RL-GNN in continuous knowledge assimilation and reduction of catastrophic forgetting.

Yeongyeong Son, Dasom Noh, Gyoungyoung Heo, Gyoung Jin Park, Sunyoung Kwon

Foundation models, serving as pretrained fundamental bases for a variety of downstream tasks, try to learn versatile, rich, and generalizable representations that can be quickly adopted through fine-tuning or even in a zero-shot manner for specific applications. Foundation models for molecular representation are no exception. Various pretext tasks have been proposed for pretraining molecular representations, but these approaches have focused on only single or partial properties. Molecules are complicated and require different perspectives depending on purposes: insights from local- or global-level, 2D-topology or 3D-spatial arrangement, and low- or high-level semantics. We propose Multi-level mOlecule gRaph prE-train (MORE) to consider these multiple aspects of molecules simultaneously. Experimental results demonstrate that our proposed method effectively learns comprehensive representations by showing outstanding performance in both linear probing and full fine-tuning. Notably, in quantification experiments of forgetting the pretrained models, MORE consistently exhibits minimal and stable parameter changes with the smallest performance gap, whereas other methods show substantial and inconsistent fluctuations with larger gaps. The effectiveness of individual pretext tasks varies depending on the problems being solved, which again highlights the need for a multi-level perspective. Scalability experiments reveal steady improvements of MORE as the dataset size increases, suggesting potential gains with larger datasets as well.

Minseok Son, Inyong Koo, Jinyoung Park, Changick Kim

When trained with severely imbalanced data, deep neural networks often struggle to accurately recognize classes with few samples. Previous studies in long-tailed recognition have attempted to rebalance biased learning using known sample distributions, primarily addressing different classification difficulties at the class level. However, these approaches often overlook the instance difficulty variation within each class. In this paper, we propose a difficulty-aware balancing margin (DBM) loss, which considers both class imbalance and instance difficulty. DBM loss comprises two components: a class-wise margin to mitigate learning bias caused by imbalanced class frequencies, and an instance-wise margin assigned to hard positive samples based on their individual difficulty. DBM loss improves class discriminativity by assigning larger margins to more difficult samples. Our method effortlessly combine with existing approaches and consistently improves performance across various long-tailed recognition benchmarks.

Akshit Sinha, Sreeram Vennam, Charu Sharma, Ponnurangam Kumaraguru

Graph Neural Networks (GNNs) have emerged as powerful tools for learning representations of graph-structured data, demonstrating remarkable performance across various tasks. Recognizing their importance, there has been extensive research focused on explaining GNN predictions, aiming to enhance their interpretability and trustworthiness. However, GNNs and their explainers face a notable challenge: graphs are primarily designed to model pair-wise relationships between nodes, which can make it tough to capture higher-order, multi-node interactions. This characteristic can pose difficulties for existing explainers in fully representing multi-node relationships. To address this gap, we present Framework For Higher-Order Representations In Graph Explanations (FORGE), a framework that enables graph explainers to capture such interactions by incorporating higher-order structures, resulting in more accurate and faithful explanations. Extensive evaluation shows that on average real-world datasets from the GraphXAI benchmark and synthetic datasets across various graph explainers, FORGE improves average explanation accuracy by 1.9x and 2.25x, respectively. We perform ablation studies to confirm the importance of higher-order relations in improving explanations, while our scalability analysis demonstrates FORGE's efficacy on large graphs.