arXiv subjects

Jinyang Li

Publications and source records attributed to Jinyang Li.

100 records · Page 6Linked to original sources

arXiv2020

Scalable Graph Neural Networks for Heterogeneous Graphs

Graph neural networks (GNNs) are a popular class of parametric model for learning over graph-structured data. Recent work has argued that GNNs primarily use the graph for feature smoothing, and have shown competitive results on benchmark tasks by simply operating on graph-smoothed node features, rather than using end-to-end learned feature hierarchies that are challenging to scale to large graphs. In this work, we ask whether these results can be extended to heterogeneous graphs, which encode multiple types of relationship between different entities. We propose Neighbor Averaging over Relation Subgraphs (NARS), which trains a classifier on neighbor-averaged features for randomly-sampled subgraphs of the "metagraph" of relations. We describe optimizations to allow these sets of node features to be computed in a memory-efficient way, both at training and inference time. NARS achieves a new state of the art accuracy on several benchmark datasets, outperforming more expensive GNN-based methods

cs.LG↗

arXiv2020

Scheduling Real-time Deep Learning Services as Imprecise Computations

The paper presents an efficient real-time scheduling algorithm for intelligent real-time edge services, defined as those that perform machine intelligence tasks, such as voice recognition, LIDAR processing, or machine vision, on behalf of local embedded devices that are themselves unable to support extensive computations. The work contributes to a recent direction in real-time computing that develops scheduling algorithms for machine intelligence tasks with anytime prediction. We show that deep neural network workflows can be cast as imprecise computations, each with a mandatory part and (several) optional parts whose execution utility depends on input data. The goal of the real-time scheduler is to maximize the average accuracy of deep neural network outputs while meeting task deadlines, thanks to opportunistic shedding of the least necessary optional parts. The work is motivated by the proliferation of increasingly ubiquitous but resource-constrained embedded devices (for applications ranging from autonomous cars to the Internet of Things) and the desire to develop services that endow them with intelligence. Experiments on recent GPU hardware and a state of the art deep neural network for machine vision illustrate that our scheme can increase the overall accuracy by 10%-20% while incurring (nearly) no deadline misses.

cs.LG↗

arXiv2020

High sensitivity multi-axes rotation sensing using large momentum transfer point source atom interferometry

A point source interferometer (PSI) is a device where atoms are split and recombined by applying a temporal sequence of Raman pulses during the expansion of a cloud of cold atoms behaving approximately as a point source. The PSI can work as a sensitive multi-axes gyroscope that can automatically filter out the signal from accelerations. The phase shift arising from rotations is proportional to the momentum transferred to each atom from the Raman pulses. Therefore, by increasing the momentum transfer, it should be possibly to enhance the sensitivity of the PSI. Here, we investigate the degree of enhancement in sensitivity that could be achieved by augmenting the PSI with large momentum transfer (LMT) employing a sequence of many Raman pulses with alternating directions. Contrary to typical approaches used for describing a PSI, we employ a model under which the motion of the center of mass of each atom is described quantum mechanically. We show how increasing Doppler shifts lead to imperfections, thereby limiting the visibility of the signal fringes, and identify ways to suppress this effect by increasing the effective, two-photon Rabi frequencies of the Raman pulses. Taking into account the effect of spontaneous emission, we show that, for a given value of the one-photon Rabi frequency, there is an optimum value for the number of pulses employed, beyond which the net enhancement in sensitivity begins to decrease. For a one-photon Rabi frequency of 200 MHz, for example, the peak value of the factor of enhancement in sensitivity is ~39, for a momentum transfer that is ~69 times as large as that for a conventional PSI. We also find that this peak value scales as the one-photon Rabi frequency to the power of 4/5.

quant-ph↗

arXiv2020

Deep Graph Library: A Graph-Centric, Highly-Performant Package for Graph Neural Networks

Advancing research in the emerging field of deep graph learning requires new tools to support tensor computation over graphs. In this paper, we present the design principles and implementation of Deep Graph Library (DGL). DGL distills the computational patterns of GNNs into a few generalized sparse tensor operations suitable for extensive parallelization. By advocating graph as the central programming abstraction, DGL can perform optimizations transparently. By cautiously adopting a framework-neutral design, DGL allows users to easily port and leverage the existing components across multiple deep learning frameworks. Our evaluation shows that DGL significantly outperforms other popular GNN-oriented frameworks in both speed and memory consumption over a variety of benchmarks and has little overhead for small scale workloads.

cs.LG↗

arXiv2020

DTG-Net: Differentiated Teachers Guided Self-Supervised Video Action Recognition

State-of-the-art video action recognition models with complex network architecture have archived significant improvements, but these models heavily depend on large-scale well-labeled datasets. To reduce such dependency, we propose a self-supervised teacher-student architecture, i.e., the Differentiated Teachers Guided self-supervised Network (DTG-Net). In DTG-Net, except for reducing labeled data dependency by self-supervised learning (SSL), pre-trained action related models are used as teacher guidance providing prior knowledge to alleviate the demand for a large number of unlabeled videos in SSL. Specifically, leveraging the years of effort in action-related tasks, e.g., image classification, image-based action recognition, the DTG-Net learns the self-supervised video representation under various teacher guidance, i.e., those well-trained models of action-related tasks. Meanwhile, the DTG-Net is optimized in the way of contrastive self-supervised learning. When two image sequences are randomly sampled from the same video or different videos as the positive or negative pairs, respectively, they are then sent to the teacher and student networks for feature embedding. After that, the contrastive feature consistency is defined between features embedding of each pair, i.e., consistent for positive pair and inconsistent for negative pairs. Meanwhile, to reflect various teacher tasks' different guidance, we also explore different weighted guidance on teacher tasks. Finally, the DTG-Net is evaluated in two ways: (i) the self-supervised DTG-Net to pre-train the supervised action recognition models with only unlabeled videos; (ii) the supervised DTG-Net to be jointly trained with the supervised action networks in an end-to-end way. Its performance is better than most pre-training methods but also has excellent competitiveness compared to supervised action recognition methods.

cs.CV↗

arXiv2020

The Paradox of Information Access: Growing Isolation in the Age of Sharing

Modern online media, such as Twitter, Instagram, and YouTube, enable anyone to become an information producer and to offer online content for potentially global consumption. By increasing the amount of globally accessible real-time information, today's ubiquitous producers contribute to a world, where an individual consumes vanishingly smaller fractions of all produced content. In general, consumers preferentially select information that closely matches their individual views and values. The bias inherent in such selection is further magnified by today's information curation services that maximize user engagement (and thus service revenue) by filtering new content in accordance with observed consumer preferences. Consequently, individuals get exposed to increasingly narrower bands of the ideology spectrum. Societies get fragmented into increasingly ideologically isolated enclaves. These enclaves (or echo-chambers) then become vulnerable to misinformation spread, which in turn further magnifies polarization and bias. We call this dynamic the paradox of information access; a growing ideological fragmentation in the age of sharing. This article describes the technical, economic, and socio-cognitive contributors to this paradox, and explores research directions towards its mitigation.

cs.CY↗

arXiv2019

On the parallels between Paxos and Raft, and how to port optimizations

In recent years, Raft has overtaken Paxos as the consensus algorithm of choice. [53] While many have pointed out similarities between the two protocols, no one has formally mapped out their relationships. In this paper, we show how Raft and Paxos are formally related despite their surface differences. Based on the formal mapping between the two protocols, we show how to automatically port a certain class of optimizations from Paxos to Raft with guaranteed correctness. As case studies, we port and evaluate two optimizations, Mencius and Paxos Quorum Lease to Raft.

cs.DC↗

arXiv2019

STFNets: Learning Sensing Signals from the Time-Frequency Perspective with Short-Time Fourier Neural Networks

Recent advances in deep learning motivate the use of deep neural networks in Internet-of-Things (IoT) applications. These networks are modelled after signal processing in the human brain, thereby leading to significant advantages at perceptual tasks such as vision and speech recognition. IoT applications, however, often measure physical phenomena, where the underlying physics (such as inertia, wireless signal propagation, or the natural frequency of oscillation) are fundamentally a function of signal frequencies, offering better features in the frequency domain. This observation leads to a fundamental question: For IoT applications, can one develop a new brand of neural network structures that synthesize features inspired not only by the biology of human perception but also by the fundamental nature of physics? Hence, in this paper, instead of using conventional building blocks (e.g., convolutional and recurrent layers), we propose a new foundational neural network building block, the Short-Time Fourier Neural Network (STFNet). It integrates a widely-used time-frequency analysis method, the Short-Time Fourier Transform, into data processing to learn features directly in the frequency domain, where the physics of underlying phenomena leave better foot-prints. STFNets bring additional flexibility to time-frequency analysis by offering novel nonlinear learnable operations that are spectral-compatible. Moreover, STFNets show that transforming signals to a domain that is more connected to the underlying physics greatly simplifies the learning process. We demonstrate the effectiveness of STFNets with extensive experiments. STFNets significantly outperform the state-of-the-art deep learning models in all experiments. A STFNet, therefore, demonstrates superior capability as the fundamental building block of deep neural networks for IoT applications for various sensor inputs.

cs.LG↗

arXiv2019

Supporting Very Large Models using Automatic Dataflow Graph Partitioning

This paper presents Tofu, a system that partitions very large DNN models across multiple GPU devices to reduce per-GPU memory footprint. Tofu is designed to partition a dataflow graph of fine-grained tensor operators in order to work transparently with a general-purpose deep learning platform like MXNet. In order to automatically partition each operator, we propose to describe the semantics of an operator in a simple language which represents tensors as lambda functions mapping from tensor coordinates to values. To optimally partition different operators in a dataflow graph, Tofu uses a recursive search algorithm that minimizes the total communication cost. Our experiments on an 8-GPU machine show that Tofu enables the training of very large CNN and RNN models. It also achieves 25% - 400% speedup over alternative approaches to train very large models.

cs.DC↗

arXiv2018

Unifying Data, Model and Hybrid Parallelism in Deep Learning via Tensor Tiling

Deep learning systems have become vital tools across many fields, but the increasing model sizes mean that training must be accelerated to maintain such systems' utility. Current systems like Tensorflow and MXNet focus on one specific parallelization strategy, data parallelism, which requires large training batch sizes in order to scale. We cast the problem of finding the best parallelization strategy as the problem of finding the best tiling to partition tensors with the least overall communication. We propose an algorithm that can find the optimal tiling. Our resulting parallelization solution is a hybrid of data parallelism and model parallelism. We build the SoyBean system that performs automatic parallelization. SoyBean automatically transforms a serial dataflow graph captured by an existing deep learning system frontend into a parallel dataflow graph based on the optimal tiling it has found. Our evaluations show that SoyBean is 1.5x-4x faster than pure data parallelism for AlexNet and VGG. We present this automatic tiling in a new system, SoyBean, that can act as a backend for Tensorflow, MXNet, and others.

cs.DC↗