Shuai Zheng

Chaoyang He, Shuai Zheng, Aston Zhang et al.

The mixture of Expert (MoE) parallelism is a recent advancement that scales up the model size with constant computational cost. MoE selects different sets of parameters (i.e., experts) for each incoming token, resulting in a sparsely-activated model. Despite several successful applications of MoE, its training efficiency degrades significantly as the number of experts increases. The routing stage in MoE relies on the efficiency of the All2All communication collective, which suffers from network congestion and has poor scalability. To mitigate these issues, we introduce SMILE, which exploits heterogeneous network bandwidth and splits a single-step routing into bi-level routing. Our experimental results show that the proposed method obtains a 2.5x speedup over Switch Transformer in terms of pretraining throughput on the Colossal Clean Crawled Corpus without losing any convergence speed.

Chenyu Jiang, Zhen Jia, Shuai Zheng et al.

Multi-task model training has been adopted to enable a single deep neural network model (often a large language model) to handle multiple tasks (e.g., question answering and text summarization). Multi-task training commonly receives input sequences of highly different lengths due to the diverse contexts of different tasks. Padding (to the same sequence length) or packing (short examples into long sequences of the same length) is usually adopted to prepare input samples for model training, which is nonetheless not space or computation efficient. This paper proposes a dynamic micro-batching approach to tackle sequence length variation and enable efficient multi-task model training. We advocate pipeline-parallel training of the large model with variable-length micro-batches, each of which potentially comprises a different number of samples. We optimize micro-batch construction using a dynamic programming-based approach, and handle micro-batch execution time variation through dynamic pipeline and communication scheduling, enabling highly efficient pipeline training. Extensive evaluation on the FLANv2 dataset demonstrates up to 4.39x higher training throughput when training T5, and 3.25x when training GPT, as compared with packing-based baselines. DynaPipe's source code is publicly available at https://github.com/awslabs/optimizing-multitask-training-through-dynamic-pipelines.

Shuai Zheng, Zhenfeng Zhu, Zhizhe Liu et al.

Graph neural networks (GNNs) have shown remarkable performance on homophilic graph data while being far less impressive when handling non-homophilic graph data due to the inherent low-pass filtering property of GNNs. In general, since real-world graphs are often complex mixtures of diverse subgraph patterns, learning a universal spectral filter on the graph from the global perspective as in most current works may still suffer from great difficulty in adapting to the variation of local patterns. On the basis of the theoretical analysis of local patterns, we rethink the existing spectral filtering methods and propose the node-oriented spectral filtering for graph neural network (namely NFGNN). By estimating the node-oriented spectral filter for each node, NFGNN is provided with the capability of precise local node positioning via the generalized translated operator, thus discriminating the variations of local homophily patterns adaptively. Meanwhile, the utilization of re-parameterization brings a good trade-off between global consistency and local sensibility for learning the node-oriented spectral filters. Furthermore, we theoretically analyze the localization property of NFGNN, demonstrating that the signal after adaptive filtering is still positioned around the corresponding node. Extensive experimental results demonstrate that the proposed NFGNN achieves more favorable performance.

Chenyu Jiang, Ye Tian, Zhen Jia et al.

The Mixture-of-Expert (MoE) technique plays a crucial role in expanding the size of DNN model parameters. However, it faces the challenge of extended all-to-all communication latency during the training process. Existing methods attempt to mitigate this issue by overlapping all-to-all with expert computation. Yet, these methods frequently fall short of achieving sufficient overlap, consequently restricting the potential for performance enhancements. In our study, we extend the scope of this challenge by considering overlap at the broader training graph level. During the forward pass, we enable non-MoE computations to overlap with all-to-all through careful partitioning and pipelining. In the backward pass, we achieve overlap with all-to-all by scheduling gradient weight computations. We implement these techniques in Lancet, a system using compiler-based optimization to automatically enhance MoE model training. Our extensive evaluation reveals that Lancet significantly reduces the time devoted to non-overlapping communication, by as much as 77%. Moreover, it achieves a notable end-to-end speedup of up to 1.3 times when compared to the state-of-the-art solutions.

Dong Chen, Shuai Zheng, Muhao Xu et al.

In the domain of dynamic graph representation learning (DGRL), the efficient and comprehensive capture of temporal evolution within real-world networks is crucial. Spiking Neural Networks (SNNs), known as their temporal dynamics and low-power characteristic, offer an efficient solution for temporal processing in DGRL task. However, owing to the spike-based information encoding mechanism of SNNs, existing DGRL methods employed SNNs face limitations in their representational capacity. Given this issue, we propose a novel framework named Spike-induced Graph Neural Network (SiGNN) for learning enhanced spatialtemporal representations on dynamic graphs. In detail, a harmonious integration of SNNs and GNNs is achieved through an innovative Temporal Activation (TA) mechanism. Benefiting from the TA mechanism, SiGNN not only effectively exploits the temporal dynamics of SNNs but also adeptly circumvents the representational constraints imposed by the binary nature of spikes. Furthermore, leveraging the inherent adaptability of SNNs, we explore an in-depth analysis of the evolutionary patterns within dynamic graphs across multiple time granularities. This approach facilitates the acquisition of a multiscale temporal node representation.Extensive experiments on various real-world dynamic graph datasets demonstrate the superior performance of SiGNN in the node classification task.

Yeyu Yan, Shuai Zheng, Wenjun Hui et al.

Graph condensation (GC) aims to distill the original graph into a small-scale graph, mitigating redundancy and accelerating GNN training. However, conventional GC approaches heavily rely on rigid GNNs and task-specific supervision. Such a dependency severely restricts their reusability and generalization across various tasks and architectures. In this work, we revisit the goal of ideal GC from the perspective of GNN optimization consistency, and then a generalized GC optimization objective is derived, by which those traditional GC methods can be viewed nicely as special cases of this optimization paradigm. Based on this, Pre-trained Graph Condensation (PreGC) via optimal transport is proposed to transcend the limitations of task- and architecture-dependent GC methods. Specifically, a hybrid-interval graph diffusion augmentation is presented to suppress the weak generalization ability of the condensed graph on particular architectures by enhancing the uncertainty of node states. Meanwhile, the matching between optimal graph transport plan and representation transport plan is tactfully established to maintain semantic consistencies across source graph and condensed graph spaces, thereby freeing graph condensation from task dependencies. To further facilitate the adaptation of condensed graphs to various downstream tasks, a traceable semantic harmonizer from source nodes to condensed nodes is proposed to bridge semantic associations through the optimized representation transport plan in pre-training. Extensive experiments verify the superiority and versatility of PreGC, demonstrating its task-independent nature and seamless compatibility with arbitrary GNNs.

Dong Chen, Shuai Zheng, Yeyu Yan et al.

Recent research on deep graph learning has shifted from static to dynamic graphs, motivated by the evolving behaviors observed in complex real-world systems. However, the temporal extension in dynamic graphs poses significant data efficiency challenges, including increased data volume, high spatiotemporal redundancy, and reliance on costly dynamic graph neural networks (DGNNs). To alleviate the concerns, we pioneer the study of dynamic graph condensation (DGC), which aims to substantially reduce the scale of dynamic graphs for data-efficient DGNN training. Accordingly, we propose DyGC, a novel framework that condenses the real dynamic graph into a compact version while faithfully preserving the inherent spatiotemporal characteristics. Specifically, to endow synthetic graphs with realistic evolving structures, a novel spiking structure generation mechanism is introduced. It draws on the dynamic behavior of spiking neurons to model temporally-aware connectivity in dynamic graphs. Given the tightly coupled spatiotemporal dependencies, DyGC proposes a tailored distribution matching approach that first constructs a semantically rich state evolving field for dynamic graphs, and then performs fine-grained spatiotemporal state alignment to guide the optimization of the condensed graph. Experiments across multiple dynamic graph datasets and representative DGNN architectures demonstrate the effectiveness of DyGC. Notably, our method retains up to 96.2% DGNN performance with only 0.5% of the original graph size, and achieves up to 1846 times training speedup.

Qiyao Wang, Ahmed Farahat, Chetan Gupta et al.

Time series data have grown at an explosive rate in numerous domains and have stimulated a surge of time series modeling research. A comprehensive comparison of different time series models, for a considered data analytics task, provides useful guidance on model selection for data analytics practitioners. Data scarcity is a universal issue that occurs in a vast range of data analytics problems, due to the high costs associated with collecting, generating, and labeling data as well as some data quality issues such as missing data. In this paper, we focus on the temporal classification/regression problem that attempts to build a mathematical mapping from multivariate time series inputs to a discrete class label or a real-valued response variable. For this specific problem, we identify two types of scarce data: scarce data with small samples and scarce data with sparsely and irregularly observed time series covariates. Observing that all existing works are incapable of utilizing the sparse time series inputs for proper modeling building, we propose a model called sparse functional multilayer perceptron (SFMLP) for handling the sparsity in the time series covariates. The effectiveness of the proposed SFMLP under each of the two types of data scarcity, in comparison with the conventional deep sequential learning models (e.g., Recurrent Neural Network, and Long Short-Term Memory), is investigated through mathematical arguments and numerical experiments.

Chi Zhang, Philip Odonkor, Shuai Zheng et al.

Dynamic dispatching is one of the core problems for operation optimization in traditional industries such as mining, as it is about how to smartly allocate the right resources to the right place at the right time. Conventionally, the industry relies on heuristics or even human intuitions which are often short-sighted and sub-optimal solutions. Leveraging the power of AI and Internet of Things (IoT), data-driven automation is reshaping this area. However, facing its own challenges such as large-scale and heterogenous trucks running in a highly dynamic environment, it can barely adopt methods developed in other domains (e.g., ride-sharing). In this paper, we propose a novel Deep Reinforcement Learning approach to solve the dynamic dispatching problem in mining. We first develop an event-based mining simulator with parameters calibrated in real mines. Then we propose an experience-sharing Deep Q Network with a novel abstract state/action representation to learn memories from heterogeneous agents altogether and realizes learning in a centralized way. We demonstrate that the proposed methods significantly outperform the most widely adopted approaches in the industry by $5.56\%$ in terms of productivity. The proposed approach has great potential in a broader range of industries (e.g., manufacturing, logistics) which have a large-scale of heterogenous equipment working in a highly dynamic environment, as a general framework for dynamic resource allocation.

Shuai Zheng, Chetan Gupta, Susumu Serita

Efficient dispatching rule in manufacturing industry is key to ensure product on-time delivery and minimum past-due and inventory cost. Manufacturing, especially in the developed world, is moving towards on-demand manufacturing meaning a high mix, low volume product mix. This requires efficient dispatching that can work in dynamic and stochastic environments, meaning it allows for quick response to new orders received and can work over a disparate set of shop floor settings. In this paper we address this problem of dispatching in manufacturing. Using reinforcement learning (RL), we propose a new design to formulate the shop floor state as a 2-D matrix, incorporate job slack time into state representation, and design lateness and tardiness rewards function for dispatching purpose. However, maintaining a separate RL model for each production line on a manufacturing shop floor is costly and often infeasible. To address this, we enhance our deep RL model with an approach for dispatching policy transfer. This increases policy generalization and saves time and cost for model training and data collection. Experiments show that: (1) our approach performs the best in terms of total discounted reward and average lateness, tardiness, (2) the proposed policy transfer approach reduces training time and increases policy generalization.

Shuai Zheng, Ahmed Farahat, Chetan Gupta

Prognostics and Health Management (PHM) is an emerging engineering discipline which is concerned with the analysis and prediction of equipment health and performance. One of the key challenges in PHM is to accurately predict impending failures in the equipment. In recent years, solutions for failure prediction have evolved from building complex physical models to the use of machine learning algorithms that leverage the data generated by the equipment. However, failure prediction problems pose a set of unique challenges that make direct application of traditional classification and prediction algorithms impractical. These challenges include the highly imbalanced training data, the extremely high cost of collecting more failure samples, and the complexity of the failure patterns. Traditional oversampling techniques will not be able to capture such complexity and accordingly result in overfitting the training data. This paper addresses these challenges by proposing a novel algorithm for failure prediction using Generative Adversarial Networks (GAN-FP). GAN-FP first utilizes two GAN networks to simultaneously generate training samples and build an inference network that can be used to predict failures for new samples. GAN-FP first adopts an infoGAN to generate realistic failure and non-failure samples, and initialize the weights of the first few layers of the inference network. The inference network is then tuned by optimizing a weighted loss objective using only real failure and non-failure samples. The inference network is further tuned using a second GAN whose purpose is to guarantee the consistency between the generated samples and corresponding labels. GAN-FP can be used for other imbalanced classification problems as well.

Shuai Zheng, Chris Ding, Feiping Nie

Singular value decomposition (SVD) is the mathematical basis of principal component analysis (PCA). Together, SVD and PCA are one of the most widely used mathematical formalism/decomposition in machine learning, data mining, pattern recognition, artificial intelligence, computer vision, signal processing, etc. In recent applications, regularization becomes an increasing trend. In this paper, we present a regularized SVD (RSVD), present an efficient computational algorithm, and provide several theoretical analysis. We show that although RSVD is non-convex, it has a closed-form global optimal solution. Finally, we apply RSVD to the application of recommender system and experimental result show that RSVD outperforms SVD significantly.

Shuai Zheng, Chris Ding

Support Vector Machine (SVM) is an efficient classification approach, which finds a hyperplane to separate data from different classes. This hyperplane is determined by support vectors. In existing SVM formulations, the objective function uses L2 norm or L1 norm on slack variables. The number of support vectors is a measure of generalization errors. In this work, we propose a Minimal SVM, which uses L0.5 norm on slack variables. The result model further reduces the number of support vectors and increases the classification performance.

Shuai Zheng, Xiao Cai, Chris Ding et al.

Real life data often includes information from different channels. For example, in computer vision, we can describe an image using different image features, such as pixel intensity, color, HOG, GIST feature, SIFT features, etc.. These different aspects of the same objects are often called multi-view (or multi-modal) data. Low-rank regression model has been proved to be an effective learning mechanism by exploring the low-rank structure of real life data. But previous low-rank regression model only works on single view data. In this paper, we propose a multi-view low-rank regression model by imposing low-rank constraints on multi-view regression model. Most importantly, we provide a closed-form solution to the multi-view low-rank regression model. Extensive experiments on 4 multi-view datasets show that the multi-view low-rank regression model outperforms single-view regression model and reveals that multi-view low-rank structure is very helpful.

Shuai Zheng, Feiping Nie, Chris Ding et al.

Linear Discriminant Analysis (LDA) is a widely-used supervised dimensionality reduction method in computer vision and pattern recognition. In null space based LDA (NLDA), a well-known LDA extension, between-class distance is maximized in the null space of the within-class scatter matrix. However, there are some limitations in NLDA. Firstly, for many data sets, null space of within-class scatter matrix does not exist, thus NLDA is not applicable to those datasets. Secondly, NLDA uses arithmetic mean of between-class distances and gives equal consideration to all between-class distances, which makes larger between-class distances can dominate the result and thus limits the performance of NLDA. In this paper, we propose a harmonic mean based Linear Discriminant Analysis, Multi-Class Discriminant Analysis (MCDA), for image classification, which minimizes the reciprocal of weighted harmonic mean of pairwise between-class distance. More importantly, MCDA gives higher priority to maximize small between-class distances. MCDA can be extended to multi-label dimension reduction. Results on 7 single-label data sets and 4 multi-label data sets show that MCDA has consistently better performance than 10 other single-label approaches and 4 other multi-label approaches in terms of classification accuracy, macro and micro average F1 score.

Shuai Zheng, Chris Ding

Kernel alignment measures the degree of similarity between two kernels. In this paper, inspired from kernel alignment, we propose a new Linear Discriminant Analysis (LDA) formulation, kernel alignment LDA (kaLDA). We first define two kernels, data kernel and class indicator kernel. The problem is to find a subspace to maximize the alignment between subspace-transformed data kernel and class indicator kernel. Surprisingly, the kernel alignment induced kaLDA objective function is very similar to classical LDA and can be expressed using between-class and total scatter matrices. This can be extended to multi-label data. We use a Stiefel-manifold gradient descent algorithm to solve this problem. We perform experiments on 8 single-label and 6 multi-label data sets. Results show that kaLDA has very good performance on many single-label and multi-label problems.