Zhejiang University
Abstract:In this paper, we strive to develop an interpretable GNNs' inference paradigm, termed MSInterpreter, which can serve as a plug-and-play scheme readily applicable to various GNNs' baselines. Unlike the most existing explanation methods, MSInterpreter provides a Message-passing Selection scheme(MSScheme) to select the critical paths for GNNs' message aggregations, which aims at reaching the self-explaination instead of post-hoc explanations. In detail, the elaborate MSScheme is designed to calculate weight factors of message aggregation paths by considering the vanilla structure and node embedding components, where the structure base aims at weight factors among node-induced substructures; on the other hand, the node embedding base focuses on weight factors via node embeddings obtained by one-layer GNN.Finally, we demonstrate the effectiveness of our approach on graph classification benchmarks.
Abstract:We investigate the problem of online learning with monotone and continuous DR-submodular reward functions, which has received great attention recently. To efficiently handle this problem, especially in the case with complicated decision sets, previous studies have proposed an efficient projection-free algorithm called Mono-Frank-Wolfe (Mono-FW) using $O(T)$ gradient evaluations and linear optimization steps in total. However, it only attains a $(1-1/e)$-regret bound of $O(T^{4/5})$. In this paper, we propose an improved projection-free algorithm, namely POBGA, which reduces the regret bound to $O(T^{3/4})$ while keeping the same computational complexity as Mono-FW. Instead of modifying Mono-FW, our key idea is to make a novel combination of a projection-based algorithm called online boosting gradient ascent, an infeasible projection technique, and a blocking technique. Furthermore, we consider the decentralized setting and develop a variant of POBGA, which not only reduces the current best regret bound of efficient projection-free algorithms for this setting from $O(T^{4/5})$ to $O(T^{3/4})$, but also reduces the total communication complexity from $O(T)$ to $O(\sqrt{T})$.
Abstract:Centralized Training with Decentralized Execution (CTDE) has recently emerged as a popular framework for cooperative Multi-Agent Reinforcement Learning (MARL), where agents can use additional global state information to guide training in a centralized way and make their own decisions only based on decentralized local policies. Despite the encouraging results achieved, CTDE makes an independence assumption on agent policies, which limits agents to adopt global cooperative information from each other during centralized training. Therefore, we argue that existing CTDE methods cannot fully utilize global information for training, leading to an inefficient joint-policy exploration and even suboptimal results. In this paper, we introduce a novel Centralized Advising and Decentralized Pruning (CADP) framework for multi-agent reinforcement learning, that not only enables an efficacious message exchange among agents during training but also guarantees the independent policies for execution. Firstly, CADP endows agents the explicit communication channel to seek and take advices from different agents for more centralized training. To further ensure the decentralized execution, we propose a smooth model pruning mechanism to progressively constraint the agent communication into a closed one without degradation in agent cooperation capability. Empirical evaluations on StarCraft II micromanagement and Google Research Football benchmarks demonstrate that the proposed framework achieves superior performance compared with the state-of-the-art counterparts. Our code will be made publicly available.
Abstract:Online convex optimization (OCO) with arbitrary delays, in which gradients or other information of functions could be arbitrarily delayed, has received increasing attention recently. Different from previous studies that focus on stationary environments, this paper investigates the delayed OCO in non-stationary environments, and aims to minimize the dynamic regret with respect to any sequence of comparators. To this end, we first propose a simple algorithm, namely DOGD, which performs a gradient descent step for each delayed gradient according to their arrival order. Despite its simplicity, our novel analysis shows that DOGD can attain an $O(\sqrt{dT}(P_T+1)$ dynamic regret bound in the worst case, where $d$ is the maximum delay, $T$ is the time horizon, and $P_T$ is the path length of comparators. More importantly, in case delays do not change the arrival order of gradients, it can automatically reduce the dynamic regret to $O(\sqrt{S}(1+P_T))$, where $S$ is the sum of delays. Furthermore, we develop an improved algorithm, which can reduce those dynamic regret bounds achieved by DOGD to $O(\sqrt{dT(P_T+1)})$ and $O(\sqrt{S(1+P_T)})$, respectively. The essential idea is to run multiple DOGD with different learning rates, and utilize a meta-algorithm to track the best one based on their delayed performance. Finally, we demonstrate that our improved algorithm is optimal in both cases by deriving a matching lower bound.
Abstract:A surge of interest has emerged in utilizing Transformers in diverse vision tasks owing to its formidable performance. However, existing approaches primarily focus on optimizing internal model architecture designs that often entail significant trial and error with high burdens. In this work, we propose a new paradigm dubbed Decision Stream Calibration that boosts the performance of general Vision Transformers. To achieve this, we shed light on the information propagation mechanism in the learning procedure by exploring the correlation between different tokens and the relevance coefficient of multiple dimensions. Upon further analysis, it was discovered that 1) the final decision is associated with tokens of foreground targets, while token features of foreground target will be transmitted into the next layer as much as possible, and the useless token features of background area will be eliminated gradually in the forward propagation. 2) Each category is solely associated with specific sparse dimensions in the tokens. Based on the discoveries mentioned above, we designed a two-stage calibration scheme, namely ViT-Calibrator, including token propagation calibration stage and dimension propagation calibration stage. Extensive experiments on commonly used datasets show that the proposed approach can achieve promising results. The source codes are given in the supplements.
Abstract:Temporal graphs exhibit dynamic interactions between nodes over continuous time, whose topologies evolve with time elapsing. The whole temporal neighborhood of nodes reveals the varying preferences of nodes. However, previous works usually generate dynamic representation with limited neighbors for simplicity, which results in both inferior performance and high latency of online inference. Therefore, in this paper, we propose a novel method of temporal graph convolution with the whole neighborhood, namely Temporal Aggregation and Propagation Graph Neural Networks (TAP-GNN). Specifically, we firstly analyze the computational complexity of the dynamic representation problem by unfolding the temporal graph in a message-passing paradigm. The expensive complexity motivates us to design the AP (aggregation and propagation) block, which significantly reduces the repeated computation of historical neighbors. The final TAP-GNN supports online inference in the graph stream scenario, which incorporates the temporal information into node embeddings with a temporal activation function and a projection layer besides several AP blocks. Experimental results on various real-life temporal networks show that our proposed TAP-GNN outperforms existing temporal graph methods by a large margin in terms of both predictive performance and online inference latency. Our code is available at \url{https://github.com/doujiang-zheng/TAP-GNN}.
Abstract:Researchers of temporal networks (e.g., social networks and transaction networks) have been interested in mining dynamic patterns of nodes from their diverse interactions. Inspired by recently powerful graph mining methods like skip-gram models and Graph Neural Networks (GNNs), existing approaches focus on generating temporal node embeddings sequentially with nodes' sequential interactions. However, the sequential modeling of previous approaches cannot handle the transition structure between nodes' neighbors with limited memorization capacity. Detailedly, an effective method for the transition structures is required to both model nodes' personalized patterns adaptively and capture node dynamics accordingly. In this paper, we propose a method, namely Transition Propagation Graph Neural Networks (TIP-GNN), to tackle the challenges of encoding nodes' transition structures. The proposed TIP-GNN focuses on the bilevel graph structure in temporal networks: besides the explicit interaction graph, a node's sequential interactions can also be constructed as a transition graph. Based on the bilevel graph, TIP-GNN further encodes transition structures by multi-step transition propagation and distills information from neighborhoods by a bilevel graph convolution. Experimental results over various temporal networks reveal the efficiency of our TIP-GNN, with at most 7.2\% improvements of accuracy on temporal link prediction. Extensive ablation studies further verify the effectiveness and limitations of the transition propagation module. Our code is available at \url{https://github.com/doujiang-zheng/TIP-GNN}.
Abstract:Vision transformers have achieved remarkable success in computer vision tasks by using multi-head self-attention modules to capture long-range dependencies within images. However, the high inference computation cost poses a new challenge. Several methods have been proposed to address this problem, mainly by slimming patches. In the inference stage, these methods classify patches into two classes, one to keep and the other to discard in multiple layers. This approach results in additional computation at every layer where patches are discarded, which hinders inference acceleration. In this study, we tackle the patch slimming problem from a different perspective by proposing a life regression module that determines the lifespan of each image patch in one go. During inference, the patch is discarded once the current layer index exceeds its life. Our proposed method avoids additional computation and parameters in multiple layers to enhance inference speed while maintaining competitive performance. Additionally, our approach requires fewer training epochs than other patch slimming methods.
Abstract:The problem of deep long-tailed learning, a prevalent challenge in the realm of generic visual recognition, persists in a multitude of real-world applications. To tackle the heavily-skewed dataset issue in long-tailed classification, prior efforts have sought to augment existing deep models with the elaborate class-balancing strategies, such as class rebalancing, data augmentation, and module improvement. Despite the encouraging performance, the limited class knowledge of the tailed classes in the training dataset still bottlenecks the performance of the existing deep models. In this paper, we propose an innovative long-tailed learning paradigm that breaks the bottleneck by guiding the learning of deep networks with external prior knowledge. This is specifically achieved by devising an elaborated ``prophetic'' teacher, termed as ``Propheter'', that aims to learn the potential class distributions. The target long-tailed prediction model is then optimized under the instruction of the well-trained ``Propheter'', such that the distributions of different classes are as distinguishable as possible from each other. Experiments on eight long-tailed benchmarks across three architectures demonstrate that the proposed prophetic paradigm acts as a promising solution to the challenge of limited class knowledge in long-tailed datasets. Our code and model can be found in the supplementary material.
Abstract:Most existing online knowledge distillation(OKD) techniques typically require sophisticated modules to produce diverse knowledge for improving students' generalization ability. In this paper, we strive to fully utilize multi-model settings instead of well-designed modules to achieve a distillation effect with excellent generalization performance. Generally, model generalization can be reflected in the flatness of the loss landscape. Since averaging parameters of multiple models can find flatter minima, we are inspired to extend the process to the sampled convex combinations of multi-student models in OKD. Specifically, by linearly weighting students' parameters in each training batch, we construct a Hybrid-Weight Model(HWM) to represent the parameters surrounding involved students. The supervision loss of HWM can estimate the landscape's curvature of the whole region around students to measure the generalization explicitly. Hence we integrate HWM's loss into students' training and propose a novel OKD framework via parameter hybridization(OKDPH) to promote flatter minima and obtain robust solutions. Considering the redundancy of parameters could lead to the collapse of HWM, we further introduce a fusion operation to keep the high similarity of students. Compared to the state-of-the-art(SOTA) OKD methods and SOTA methods of seeking flat minima, our OKDPH achieves higher performance with fewer parameters, benefiting OKD with lightweight and robust characteristics. Our code is publicly available at https://github.com/tianlizhang/OKDPH.