Search

Graph-based patterns are extensively employed and favored by practitioners within industrial companies due to their capacity to represent the behavioral attributes and topological relationships among users, thereby offering enhanced interpretability in comparison to black-box models commonly utilized for classification and recognition tasks. For instance, within the scenario of transaction risk management, a graph pattern that is characteristic of a particular risk category can be readily employed to discern transactions fraught with risk, delineate networks of criminal activity, or investigate the methodologies employed by fraudsters. Nonetheless, graph data in industrial settings is often characterized by its massive scale, encompassing data sets with millions or even billions of nodes, making the manual extraction of graph patterns not only labor-intensive but also necessitating specialized knowledge in particular domains of risk. Moreover, existing methodologies for mining graph patterns encounter significant obstacles when tasked with analyzing large-scale attributed graphs. In this work, we introduce GraphRPM, an industry-purpose parallel and distributed risk pattern mining framework on large attributed graphs. The framework incorporates a novel edge-involved graph isomorphism network alongside optimized operations for parallel graph computation, which collectively contribute to a considerable reduction in computational complexity and resource expenditure. Moreover, the intelligent filtration of efficacious risky graph patterns is facilitated by the proposed evaluation metrics. Comprehensive experimental evaluations conducted on real-world datasets of varying sizes substantiate the capability of GraphRPM to adeptly address the challenges inherent in mining patterns from large-scale industrial attributed graphs, thereby underscoring its substantial value for industrial deployment., Comment: Accepted by ECML PKDD 2024

Commonsense question answering is a crucial task that requires machines to employ reasoning according to commonsense. Previous studies predominantly employ an extracting-and-modeling paradigm to harness the information in KG, which first extracts relevant subgraphs based on pre-defined rules and then proceeds to design various strategies aiming to improve the representations and fusion of the extracted structural knowledge. Despite their effectiveness, there are still two challenges. On one hand, subgraphs extracted by rule-based methods may have the potential to overlook critical nodes and result in uncontrollable subgraph size. On the other hand, the misalignment between graph and text modalities undermines the effectiveness of knowledge fusion, ultimately impacting the task performance. To deal with the problems above, we propose a novel framework: \textbf{S}ubgraph R\textbf{E}trieval Enhanced by Gra\textbf{P}h-\textbf{T}ext \textbf{A}lignment, named \textbf{SEPTA}. Firstly, we transform the knowledge graph into a database of subgraph vectors and propose a BFS-style subgraph sampling strategy to avoid information loss, leveraging the analogy between BFS and the message-passing mechanism. In addition, we propose a bidirectional contrastive learning approach for graph-text alignment, which effectively enhances both subgraph retrieval and knowledge fusion. Finally, all the retrieved information is combined for reasoning in the prediction module. Extensive experiments on five datasets demonstrate the effectiveness and robustness of our framework., Comment: Accepted by ECML PKDD 2024

Recently, research on Text-Attributed Graphs (TAGs) has gained significant attention due to the prevalence of free-text node features in real-world applications and the advancements in Large Language Models (LLMs) that bolster TAG methodologies. However, current TAG approaches face two primary challenges: (i) Heavy reliance on label information and (ii) Limited cross-domain zero/few-shot transferability. These issues constrain the scaling of both data and model size, owing to high labor costs and scaling laws, complicating the development of graph foundation models with strong transferability. In this work, we propose the GraphCLIP framework to address these challenges by learning graph foundation models with strong cross-domain zero/few-shot transferability through a self-supervised contrastive graph-summary pretraining method. Specifically, we generate and curate large-scale graph-summary pair data with the assistance of LLMs, and introduce a novel graph-summary pretraining method, combined with invariant learning, to enhance graph foundation models with strong cross-domain zero-shot transferability. For few-shot learning, we propose a novel graph prompt tuning technique aligned with our pretraining objective to mitigate catastrophic forgetting and minimize learning costs. Extensive experiments show the superiority of GraphCLIP in both zero-shot and few-shot settings, while evaluations across various downstream tasks confirm the versatility of GraphCLIP. Our code is available at: https://github.com/ZhuYun97/GraphCLIP, Comment: Under Review

Recent research has explored the use of Large Language Models (LLMs) for tackling complex graph reasoning tasks. However, due to the intricacies of graph structures and the inherent limitations of LLMs in handling long text, current approaches often fail to deliver satisfactory accuracy, even on small-scale graphs and simple tasks. To address these challenges, we introduce GraphAgent-Reasoner, a fine-tuning-free framework that utilizes a multi-agent collaboration strategy for explicit and precise graph reasoning. Inspired by distributed graph computation theory, our framework decomposes graph problems into smaller, node-centric tasks that are distributed among multiple agents. The agents collaborate to solve the overall problem, significantly reducing the amount of information and complexity handled by a single LLM, thus enhancing the accuracy of graph reasoning. By simply increasing the number of agents, GraphAgent-Reasoner can efficiently scale to accommodate larger graphs with over 1,000 nodes. Evaluated on the GraphInstruct dataset, our framework demonstrates near-perfect accuracy on polynomial-time graph reasoning tasks, significantly outperforming the best available models, both closed-source and fine-tuned open-source variants. Our framework also demonstrates the capability to handle real-world graph reasoning applications such as webpage importance analysis.

Knowledge graph embedding (KGE) constitutes a foundational task, directed towards learning representations for entities and relations within knowledge graphs (KGs), with the objective of crafting representations comprehensive enough to approximate the logical and symbolic interconnections among entities. In this paper, we define a metric Z-counts to measure the difficulty of training each triple ($<$head entity, relation, tail entity$>$) in KGs with theoretical analysis. Based on this metric, we propose \textbf{CL4KGE}, an efficient \textbf{C}urriculum \textbf{L}earning based training strategy for \textbf{KGE}. This method includes a difficulty measurer and a training scheduler that aids in the training of KGE models. Our approach possesses the flexibility to act as a plugin within a wide range of KGE models, with the added advantage of adaptability to the majority of KGs in existence. The proposed method has been evaluated on popular KGE models, and the results demonstrate that it enhances the state-of-the-art methods. The use of Z-counts as a metric has enabled the identification of challenging triples in KGs, which helps in devising effective training strategies., Comment: 16 pages, 3 figures

Recently, Retrieval-Augmented Generation (RAG) has achieved remarkable success in addressing the challenges of Large Language Models (LLMs) without necessitating retraining. By referencing an external knowledge base, RAG refines LLM outputs, effectively mitigating issues such as ``hallucination'', lack of domain-specific knowledge, and outdated information. However, the complex structure of relationships among different entities in databases presents challenges for RAG systems. In response, GraphRAG leverages structural information across entities to enable more precise and comprehensive retrieval, capturing relational knowledge and facilitating more accurate, context-aware responses. Given the novelty and potential of GraphRAG, a systematic review of current technologies is imperative. This paper provides the first comprehensive overview of GraphRAG methodologies. We formalize the GraphRAG workflow, encompassing Graph-Based Indexing, Graph-Guided Retrieval, and Graph-Enhanced Generation. We then outline the core technologies and training methods at each stage. Additionally, we examine downstream tasks, application domains, evaluation methodologies, and industrial use cases of GraphRAG. Finally, we explore future research directions to inspire further inquiries and advance progress in the field. In order to track recent progress in this field, we set up a repository at \url{https://github.com/pengboci/GraphRAG-Survey}., Comment: Ongoing work. Compared to the first version, several references have been added and a GitHub repository link has been provided

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

In the graph domain, deep graph networks based on Message Passing Neural Networks (MPNNs) or Graph Transformers often cause over-smoothing of node features, limiting their expressive capacity. Many upsampling techniques involving node and edge manipulation have been proposed to mitigate this issue. However, these methods often require extensive manual labor, resulting in suboptimal performance and lacking a universal integration strategy. In this study, we introduce UniGAP, a universal and adaptive graph upsampling technique for graph data. It provides a universal framework for graph upsampling, encompassing most current methods as variants. Moreover, UniGAP serves as a plug-in component that can be seamlessly and adaptively integrated with existing GNNs to enhance performance and mitigate the over-smoothing problem. Through extensive experiments, UniGAP demonstrates significant improvements over heuristic data augmentation methods across various datasets and metrics. We analyze how graph structure evolves with UniGAP, identifying key bottlenecks where over-smoothing occurs, and providing insights into how UniGAP addresses this issue. Lastly, we show the potential of combining UniGAP with large language models (LLMs) to further improve downstream performance. Our code is available at: https://github.com/wangxiaotang0906/UniGAP

Adaptive optimizers, such as Adam, have achieved remarkable success in deep learning. A key component of these optimizers is the so-called preconditioning matrix, providing enhanced gradient information and regulating the step size of each gradient direction. In this paper, we propose a novel approach to designing the preconditioning matrix by utilizing the gradient difference between two successive steps as the diagonal elements. These diagonal elements are closely related to the Hessian and can be perceived as an approximation of the inner product between the Hessian row vectors and difference of the adjacent parameter vectors. Additionally, we introduce an auto-switching function that enables the preconditioning matrix to switch dynamically between Stochastic Gradient Descent (SGD) and the adaptive optimizer. Based on these two techniques, we develop a new optimizer named AGD that enhances the generalization performance. We evaluate AGD on public datasets of Natural Language Processing (NLP), Computer Vision (CV), and Recommendation Systems (RecSys). Our experimental results demonstrate that AGD outperforms the state-of-the-art (SOTA) optimizers, achieving highly competitive or significantly better predictive performance. Furthermore, we analyze how AGD is able to switch automatically between SGD and the adaptive optimizer and its actual effects on various scenarios. The code is available at https://github.com/intelligent-machine-learning/dlrover/tree/master/atorch/atorch/optimizers., Comment: 21 pages. Accepted as a conference paper at NeurIPS '23

Due to 5G deployment, there is significant interest in LDPC decoding. While much research is devoted on efficient hardwiring of algorithms based on Belief Propagation (BP), it has been shown that LDPC decoding can be formulated as a combinatorial optimization problem, which could benefit from significant acceleration of physical computation mechanisms such as Ising machines. This approach has so far resulted in poor performance. This paper shows that the reason is not fundamental but suboptimal hardware and formulation. A co-designed Ising machine-based system can improve speed by 3 orders of magnitude. As a result, a physical computation approach can outperform hardwiring state-of-the-art algorithms. In this paper, we show such an augmented Ising machine that is 4.4$\times$ more energy efficient than the state of the art in the literature.

This paper studies the distributed optimization problem over directed networks with noisy information-sharing. To resolve the imperfect communication issue over directed networks, a series of noise-robust variants of Push-Pull/AB method have been developed. These methods improve the robustness of Push-Pull method against the information-sharing noise through adding small factors on weight matrices and replacing the global gradient tracking with the cumulative gradient tracking. Based on the two techniques, we propose a new variant of the Push-Pull method by presenting a novel mechanism of inter-agent information aggregation, named variance-reduced aggregation (VRA). VRA helps us to release some conditions on the objective function and networks. When the objective function is convex and the sharing-information noise is variance-unbounded, it can be shown that the proposed method converges to the optimal solution almost surely. When the objective function is strongly convex and the sharing-information noise is variance-bounded, the proposed method achieves the convergence rate of $\mathcal{O}\left(k^{-(1-\epsilon)}\right)$ in the mean square sense, where $\epsilon$ could be close to 0 infinitely. Simulated experiments on ridge regression problems verify the effectiveness of the proposed method.

Deep Neural Networks (DNNs) generalization is known to be closely related to the flatness of minima, leading to the development of Sharpness-Aware Minimization (SAM) for seeking flatter minima and better generalization. In this paper, we revisit the loss of SAM and propose a more general method, called WSAM, by incorporating sharpness as a regularization term. We prove its generalization bound through the combination of PAC and Bayes-PAC techniques, and evaluate its performance on various public datasets. The results demonstrate that WSAM achieves improved generalization, or is at least highly competitive, compared to the vanilla optimizer, SAM and its variants. The code is available at https://github.com/intelligent-machine-learning/dlrover/tree/master/atorch/atorch/optimizers., Comment: 10 pages. Accepted as a conference paper at KDD '23

Nature apparently does a lot of computation constantly. If we can harness some of that computation at an appropriate level, we can potentially perform certain type of computation (much) faster and more efficiently than we can do with a von Neumann computer. Indeed, many powerful algorithms are inspired by nature and are thus prime candidates for nature-based computation. One particular branch of this effort that has seen some recent rapid advances is Ising machines. Some Ising machines are already showing better performance and energy efficiency for optimization problems. Through design iterations and co-evolution between hardware and algorithm, we expect more benefits from nature-based computing systems. In this paper, we make a case for an augmented Ising machine suitable for both training and inference using an energy-based machine learning algorithm. We show that with a small change, the Ising substrate accelerate key parts of the algorithm and achieve non-trivial speedup and efficiency gain. With a more substantial change, we can turn the machine into a self-sufficient gradient follower to virtually complete training entirely in hardware. This can bring about 29x speedup and about 1000x reduction in energy compared to a Tensor Processing Unit (TPU) host.

The paper studies the distributed stochastic compositional optimization problems over networks, where all the agents' inner-level function is the sum of each agent's private expectation function. Focusing on the aggregative structure of the inner-level function, we employ the hybrid variance reduction method to obtain the information on each agent's private expectation function, and apply the dynamic consensus mechanism to track the information on each agent's inner-level function. Then by combining with the standard distributed stochastic gradient descent method, we propose a distributed aggregative stochastic compositional gradient descent method. When the objective function is smooth, the proposed method achieves the optimal convergence rate $\mathcal{O}\left(K^{-1/2}\right)$. We further combine the proposed method with the communication compression and propose the communication compressed variant distributed aggregative stochastic compositional gradient descent method. The compressed variant of the proposed method maintains the optimal convergence rate $\mathcal{O}\left(K^{-1/2}\right)$. Simulated experiments on decentralized reinforcement learning verify the effectiveness of the proposed methods.

We study the distributed stochastic compositional optimization problems over directed communication networks in which agents privately own a stochastic compositional objective function and collaborate to minimize the sum of all objective functions. We propose a distributed stochastic compositional gradient descent method, where the gradient tracking and the stochastic correction techniques are employed to adapt to the networks' directed structure and increase the accuracy of inner function estimation. When the objective function is smooth, the proposed method achieves the convergence rate $\mathcal{O}\left(k^{-1/2}\right)$ and sample complexity $\mathcal{O}\left(\frac{1}{\epsilon^2}\right)$ for finding the ($\epsilon$)-stationary point. When the objective function is strongly convex, the convergence rate is improved to $\mathcal{O}\left(k^{-1}\right)$. Moreover, the asymptotic normality of Polyak-Ruppert averaged iterates of the proposed method is also presented. We demonstrate the empirical performance of the proposed method on model-agnostic meta-learning problem and logistic regression problem.

The sample average approximation (SAA) and the stochastic approximation (SA) are two popular schemes for solving the stochastic variational inequalities problem (SVIP). In the past decades, theories on the consistency of the SAA solutions and SA solutions have been well studied. More recently, the asymptotic confidence regions of the true solution to SVIP have been constructed when the SAA scheme is implemented. It is of fundamental interest to develop confidence regions of the true solution to the SVIP when the SA scheme is employed. In this paper, we discuss the framework of constructing asymptotic confidence regions for the true solution of SVIP with a focus on stochastic dual average method. We first establish the asymptotic normality of the SA solutions both in ergodic sense and non-ergodic sense. Then the online methods of estimating the covariance matrices in the normal distributions are studied. Finally, practical procedures of building the asymptotic confidence regions of solutions to SVIP with numerical simulations are presented.

The popular $\mathcal{AB}$/push-pull method for distributed optimization problem may unify much of the existing decentralized first-order methods based on gradient tracking technique. More recently, the stochastic gradient variant of $\mathcal{AB}$/Push-Pull method ($\mathcal{S}$-$\mathcal{AB}$) has been proposed, which achieves the linear rate of converging to a neighborhood of the global minimizer when the step-size is constant. This paper is devoted to the asymptotic properties of $\mathcal{S}$-$\mathcal{AB}$ with diminishing stepsize. Specifically, under the condition that each local objective is smooth and the global objective is strongly-convex, we first present the boundedness of the iterates of $\mathcal{S}$-$\mathcal{AB}$ and then show that the iterates converge to the global minimizer with the rate $\mathcal{O}\left(1/\sqrt{k}\right)$. Furthermore, the asymptotic normality of Polyak-Ruppert averaged $\mathcal{S}$-$\mathcal{AB}$ is obtained and applications on statistical inference are discussed. Finally, numerical tests are conducted to demonstrate the theoretic results.

We develop a novel framework that adds the regularizers of the sparse group lasso to a family of adaptive optimizers in deep learning, such as Momentum, Adagrad, Adam, AMSGrad, AdaHessian, and create a new class of optimizers, which are named Group Momentum, Group Adagrad, Group Adam, Group AMSGrad and Group AdaHessian, etc., accordingly. We establish theoretically proven convergence guarantees in the stochastic convex settings, based on primal-dual methods. We evaluate the regularized effect of our new optimizers on three large-scale real-world ad click datasets with state-of-the-art deep learning models. The experimental results reveal that compared with the original optimizers with the post-processing procedure which uses the magnitude pruning method, the performance of the models can be significantly improved on the same sparsity level. Furthermore, in comparison to the cases without magnitude pruning, our methods can achieve extremely high sparsity with significantly better or highly competitive performance. The code is available at https://github.com/intelligent-machine-learning/dlrover/blob/master/tfplus., Comment: 24 pages. Published as a conference paper at ECML PKDD 2021. This version includes Appendix which was not included in the published version because of page limit

The large-scale recommender system mainly consists of two stages: matching and ranking. The matching stage (also known as the retrieval step) identifies a small fraction of relevant items from billion-scale item corpus in low latency and computational cost. Item-to-item collaborative filter (item-based CF) and embedding-based retrieval (EBR) have been long used in the industrial matching stage owing to its efficiency. However, item-based CF is hard to meet personalization, while EBR has difficulty in satisfying diversity. In this paper, we propose a novel matching architecture, Path-based Deep Network (named PDN), which can incorporate both personalization and diversity to enhance matching performance. Specifically, PDN is comprised of two modules: Trigger Net and Similarity Net. PDN utilizes Trigger Net to capture the user's interest in each of his/her interacted item, and Similarity Net to evaluate the similarity between each interacted item and the target item based on these items' profile and CF information. The final relevance between the user and the target item is calculated by explicitly considering user's diverse interests, \ie aggregating the relevance weights of the related two-hop paths (one hop of a path corresponds to user-item interaction and the other to item-item relevance). Furthermore, we describe the architecture design of a matching system with the proposed PDN in a leading real-world E-Commerce service (Mobile Taobao App). Based on offline evaluations and online A/B test, we show that PDN outperforms the existing solutions for the same task. The online results also demonstrate that PDN can retrieve more personalized and more diverse relevant items to significantly improve user engagement. Currently, PDN system has been successfully deployed at Mobile Taobao App and handling major online traffic., Comment: Accepted for publication in SIGIR2021

Graph neural networks (GNNs) have been popularly used in analyzing graph-structured data, showing promising results in various applications such as node classification, link prediction and network recommendation. In this paper, we present a new graph attention neural network, namely GIPA, for attributed graph data learning. GIPA consists of three key components: attention, feature propagation and aggregation. Specifically, the attention component introduces a new multi-layer perceptron based multi-head to generate better non-linear feature mapping and representation than conventional implementations such as dot-product. The propagation component considers not only node features but also edge features, which differs from existing GNNs that merely consider node features. The aggregation component uses a residual connection to generate the final embedding. We evaluate the performance of GIPA using the Open Graph Benchmark proteins (ogbn-proteins for short) dataset. The experimental results reveal that GIPA can beat the state-of-the-art models in terms of prediction accuracy, e.g., GIPA achieves an average test ROC-AUC of $0.8700\pm 0.0010$ and outperforms all the previous methods listed in the ogbn-proteins leaderboard., Comment: 5 pages, 2 figures, 2 tables; Technical report

Graph neural networks (GNNs) have been demonstrated as a powerful tool for analyzing non-Euclidean graph data. However, the lack of efficient distributed graph learning systems severely hinders applications of GNNs, especially when graphs are big and GNNs are relatively deep. Herein, we present GraphTheta, the first distributed and scalable graph learning system built upon vertex-centric distributed graph processing with neural network operators implemented as user-defined functions. This system supports multiple training strategies and enables efficient and scalable big-graph learning on distributed (virtual) machines with low memory. To facilitate graph convolutions, GraphTheta puts forward a new graph learning abstraction named NN-TGAR to bridge the gap between graph processing and graph deep learning. A distributed graph engine is proposed to conduct the stochastic gradient descent optimization with a hybrid-parallel execution, and a new cluster-batched training strategy is supported. We evaluate GraphTheta using several datasets with network sizes ranging from small-, modest- to large-scale. Experimental results show that GraphTheta can scale well to 1,024 workers for training an in-house developed GNN on an industry-scale Alipay dataset of 1.4 billion nodes and 4.1 billion attributed edges, with a cluster of CPU virtual machines (dockers) of small memory each (5$\sim$12GB). Moreover, GraphTheta can outperform DistDGL by up to $2.02\times$, with better scalability, and GraphLearn by up to $30.56\times$. As for model accuracy, GraphTheta is capable of learning as good GNNs as existing frameworks. To the best of our knowledge, this work presents the largest edge-attributed GNN learning task in the literature., Comment: 19 pages, 13 figures, 8 tables

Considering the constrained stochastic optimization problem over a time-varying random network, where the agents are to collectively minimize a sum of objective functions subject to a common constraint set, we investigate asymptotic properties of a distributed algorithm based on dual averaging of gradients. Different from most existing works on distributed dual averaging algorithms that mainly concentrating on their non-asymptotic properties, we not only prove almost sure convergence and the rate of almost sure convergence, but also asymptotic normality and asymptotic efficiency of the algorithm. Firstly, for general constrained convex optimization problem distributed over a random network, we prove that almost sure consensus can be archived and the estimates of agents converge to the same optimal point. For the case of linear constrained convex optimization, we show that the mirror map of the averaged dual sequence identifies the active constraints of the optimal solution with probability 1, which helps us to prove the almost sure convergence rate and then establish asymptotic normality of the algorithm. Furthermore, we also verify that the algorithm is asymptotically optimal. To the best of our knowledge, it seems to be the first asymptotic normality result for constrained distributed optimization algorithms. Finally, a numerical example is provided to justify the theoretical analysis.

Accelerating deep model training and inference is crucial in practice. Existing deep learning frameworks usually concentrate on optimizing training speed and pay fewer attentions to inference-specific optimizations. Actually, model inference differs from training in terms of computation, e.g. parameters are refreshed each gradient update step during training, but kept invariant during inference. These special characteristics of model inference open new opportunities for its optimization. In this paper, we propose a hardware-aware optimization framework, namely Woodpecker-DL (WPK), to accelerate inference by taking advantage of multiple joint optimizations from the perspectives of graph optimization, automated searches, domain-specific language (DSL) compiler techniques and system-level exploration. In WPK, we investigated two new automated search approaches based on genetic algorithm and reinforcement learning, respectively, to hunt the best operator code configurations targeting specific hardware. A customized DSL compiler is further attached to these search algorithms to generate efficient codes. To create an optimized inference plan, WPK systematically explores high-speed operator implementations from third-party libraries besides our automatically generated codes and singles out the best implementation per operator for use. Extensive experiments demonstrated that on a Tesla P100 GPU, we can achieve the maximum speedup of 5.40 over cuDNN and 1.63 over TVM on individual convolution operators, and run up to 1.18 times faster than TensorRT for end-to-end model inference., Comment: 10 pages, 3 figures

Gaze estimation for ordinary smart phone, e.g. estimating where the user is looking at on the phone screen, can be applied in various applications. However, the widely used appearance-based CNN methods still have two issues for practical adoption. First, due to the limited dataset, gaze estimation is very likely to suffer from over-fitting, leading to poor accuracy at run time. Second, the current methods are usually not robust, i.e. their prediction results having notable jitters even when the user is performing gaze fixation, which degrades user experience greatly. For the first issue, we propose a new tolerant and talented (TAT) training scheme, which is an iterative random knowledge distillation framework enhanced with cosine similarity pruning and aligned orthogonal initialization. The knowledge distillation is a tolerant teaching process providing diverse and informative supervision. The enhanced pruning and initialization is a talented learning process prompting the network to escape from the local minima and re-born from a better start. For the second issue, we define a new metric to measure the robustness of gaze estimator, and propose an adversarial training based Disturbance with Ordinal loss (DwO) method to improve it. The experimental results show that our TAT method achieves state-of-the-art performance on GazeCapture dataset, and that our DwO method improves the robustness while keeping comparable accuracy., Comment: Accepted by ICCV 2019 Workshop. Fix the error of the Figure 1 in the camera ready file

Pairwise association measure is an important operation in data analytics. Kendall's tau coefficient is one widely used correlation coefficient identifying non-linear relationships between ordinal variables. In this paper, we investigated a parallel algorithm accelerating all-pairs Kendall's tau coefficient computation via single instruction multiple data (SIMD) vectorized sorting on Intel Xeon Phis by taking advantage of many processing cores and 512-bit SIMD vector instructions. To facilitate workload balancing and overcome on-chip memory limitation, we proposed a generic framework for symmetric all-pairs computation by building provable bijective functions between job identifier and coordinate space. Performance evaluation demonstrated that our algorithm on one 5110P Phi achieves two orders-of-magnitude speedups over 16-threaded MATLAB and three orders-of-magnitude speedups over sequential R, both running on high-end CPUs. Besides, our algorithm exhibited rather good distributed computing scalability with respect to number of Phis. Source code and datasets are publicly available at http://lightpcc.sourceforge.net., Comment: 29 pages, 6 figures, 5 tables, submitted to Journal of Parallel and Distributed Computing

Co-expression network is a critical technique for the identification of inter-gene interactions, which usually relies on all-pairs correlation (or similar measure) computation between gene expression profiles across multiple samples. Pearson's correlation coefficient (PCC) is one widely used technique for gene co-expression network construction. However, all-pairs PCC computation is computationally demanding for large numbers of gene expression profiles, thus motivating our acceleration of its execution using high-performance computing. In this paper, we present LightPCC, the first parallel and distributed all-pairs PCC computation on Intel Xeon Phi (Phi) clusters. It achieves high speed by exploring the SIMD-instruction-level and thread-level parallelism within Phis as well as accelerator-level parallelism among multiple Phis. To facilitate balanced workload distribution, we have proposed a general framework for symmetric all-pairs computation by building bijective functions between job identifier and coordinate space for the first time. We have evaluated LightPCC and compared it to two CPU-based counterparts: a sequential C++ implementation in ALGLIB and an implementation based on a parallel general matrix-matrix multiplication routine in Intel Math Kernel Library (MKL) (all use double precision), using a set of gene expression datasets. Performance evaluation revealed that with one 5110P Phi and 16 Phis, LightPCC runs up to $20.6\times$ and $218.2\times$ faster than ALGLIB, and up to $6.8\times$ and $71.4\times$ faster than single-threaded MKL, respectively. In addition, LightPCC demonstrated good parallel scalability in terms of number of Phis. Source code of LightPCC is publicly available at http://lightpcc.sourceforge.net., Comment: 9 pages, 2 figures, 2 tables, accepted by the SBAC-PAD 2016 conference

Scan (or prefix sum) is a fundamental and widely used primitive in parallel computing. In this paper, we present LightScan, a faster parallel scan primitive for CUDA-enabled GPUs, which investigates a hybrid model combining intra-block computation and inter-block communication to perform a scan. Our algorithm employs warp shuffle functions to implement fast intra-block computation and takes advantage of globally coherent L2 cache and the associated parallel thread execution (PTX) assembly instructions to realize lightweight inter-block communication. Performance evaluation using a single Tesla K40c GPU shows that LightScan outperforms existing GPU algorithms and implementations, and yields a speedup of up to 2.1, 2.4, 1.5 and 1.2 over the leading CUDPP, Thrust, ModernGPU and CUB implementations running on the same GPU, respectively. Furthermore, LightScan runs up to 8.9 and 257.3 times faster than Intel TBB running on 16 CPU cores and an Intel Xeon Phi 5110P coprocessor, respectively. Source code of LightScan is available at http://cupbb.sourceforge.net., Comment: 21 pages, 6 figures, submitted to Journal of Parallel and Distributed Computing on Jan 09, 2016

The maximal sensitivity of the Smith-Waterman (SW) algorithm has enabled its wide use in biological sequence database search. Unfortunately, the high sensitivity comes at the expense of quadratic time complexity, which makes the algorithm computationally demanding for big databases. In this paper, we present SWAPHI, the first parallelized algorithm employing Xeon Phi coprocessors to accelerate SW protein database search. SWAPHI is designed based on the scale-and-vectorize approach, i.e. it boosts alignment speed by effectively utilizing both the coarse-grained parallelism from the many co-processing cores (scale) and the fine-grained parallelism from the 512-bit wide single instruction, multiple data (SIMD) vectors within each core (vectorize). By searching against the large UniProtKB/TrEMBL protein database, SWAPHI achieves a performance of up to 58.8 billion cell updates per second (GCUPS) on one coprocessor and up to 228.4 GCUPS on four coprocessors. Furthermore, it demonstrates good parallel scalability on varying number of coprocessors, and is also superior to both SWIPE on 16 high-end CPU cores and BLAST+ on 8 cores when using four coprocessors, with the maximum speedup of 1.52 and 1.86, respectively. SWAPHI is written in C++ language (with a set of SIMD intrinsics), and is freely available at http://swaphi.sourceforge.net., Comment: A short version of this paper has been accepted by the IEEE ASAP 2014 conference

Summary: We present an extension of CUSHAW2 for fast and accurate alignments of SOLiD color-space short-reads. Our extension introduces a double-seeding approach to improve mapping sensitivity, by combining maximal exact match seeds and variable-length seeds derived from local alignments. We have compared the performance of CUSHAW2 to SHRiMP2 and BFAST by aligning both simulated and real color-space mate-paired reads to the human genome. The results show that CUSHAW2 achieves comparable or better alignment quality compared to SHRiMP2 and BFAST at an order-of-magnitude faster speed and significantly smaller peak resident memory size. Availability: CUSHAW2 and all simulated datasets are available at http://cushaw2.sourceforge.net. Contact: liuy@uni-mainz.de; bertil.schmidt@uni-mainz.de, Comment: 2 pages,1 table