arXiv subjects

Xiang He

Publications and source records attributed to Xiang He.

At least 55 records · Page 3Linked to original sources

arXiv2023

Riemannian metrics with prescribed volume and finite parts of Dirichlet spectrum

In this paper we study the problem of prescribing Dirichlet eigenvalues on an arbitrary compact manifold $M$ of dimension $n\geq 3$ with a non-empty smooth boundary $\partial M$. We show that for any finite increasing sequence of real numbers $0<a_1<a_2 \leq a_3 \leq \cdots \leq a_N$ and any positive number $V$, there exists a Riemannian metric $g$ on $M$ such that $\mathrm{Vol}(M,g)=V$ and $λ^\mathcal{D}_k(M,g)=a_k$ for any integer $1 \leq k \leq N$.

math.DG↗

arXiv2023

A Survey on Deep Neural Network Partition over Cloud, Edge and End Devices

Deep neural network (DNN) partition is a research problem that involves splitting a DNN into multiple parts and offloading them to specific locations. Because of the recent advancement in multi-access edge computing and edge intelligence, DNN partition has been considered as a powerful tool for improving DNN inference performance when the computing resources of edge and end devices are limited and the remote transmission of data from these devices to clouds is costly. This paper provides a comprehensive survey on the recent advances and challenges in DNN partition approaches over the cloud, edge, and end devices based on a detailed literature collection. We review how DNN partition works in various application scenarios, and provide a unified mathematical model of the DNN partition problem. We developed a five-dimensional classification framework for DNN partition approaches, consisting of deployment locations, partition granularity, partition constraints, optimization objectives, and optimization algorithms. Each existing DNN partition approache can be perfectly defined in this framework by instantiating each dimension into specific values. In addition, we suggest a set of metrics for comparing and evaluating the DNN partition approaches. Based on this, we identify and discuss research challenges that have not yet been investigated or fully addressed. We hope that this work helps DNN partition researchers by highlighting significant future research directions in this domain.

cs.DC↗

arXiv2023

MSAT: Biologically Inspired Multi-Stage Adaptive Threshold for Conversion of Spiking Neural Networks

Spiking Neural Networks (SNNs) can do inference with low power consumption due to their spike sparsity. ANN-SNN conversion is an efficient way to achieve deep SNNs by converting well-trained Artificial Neural Networks (ANNs). However, the existing methods commonly use constant threshold for conversion, which prevents neurons from rapidly delivering spikes to deeper layers and causes high time delay. In addition, the same response for different inputs may result in information loss during the information transmission. Inspired by the biological model mechanism, we propose a multi-stage adaptive threshold (MSAT). Specifically, for each neuron, the dynamic threshold varies with firing history and input properties and is positively correlated with the average membrane potential and negatively correlated with the rate of depolarization. The self-adaptation to membrane potential and input allows a timely adjustment of the threshold to fire spike faster and transmit more information. Moreover, we analyze the Spikes of Inactivated Neurons error which is pervasive in early time steps and propose spike confidence accordingly as a measurement of confidence about the neurons that correctly deliver spikes. We use such spike confidence in early time steps to determine whether to elicit spike to alleviate this error. Combined with the proposed method, we examine the performance on non-trivial datasets CIFAR-10, CIFAR-100, and ImageNet. We also conduct sentiment classification and speech recognition experiments on the IDBM and Google speech commands datasets respectively. Experiments show near-lossless and lower latency ANN-SNN conversion. To the best of our knowledge, this is the first time to build a biologically inspired multi-stage adaptive threshold for converted SNN, with comparable performance to state-of-the-art methods while improving energy efficiency.

cs.NE↗

arXiv2023

On the Severi problem in arbitrary characteristic

In this paper, we show that Severi varieties parameterizing irreducible reduced planar curves of a given degree and geometric genus are either empty or irreducible in any characteristic. Following Severi's original idea, this gives a new proof of the irreducibility of the moduli space of smooth projective curves of a given genus in positive characteristic. It is the first proof that involves no reduction to the characteristic zero case. As a further consequence, we generalize Zariski's theorem to positive characteristic and show that a general reduced planar curve of a given geometric genus is nodal.

math.AG↗

arXiv2022

RescureService: A Benchmark Microservice System for the Research of Mobile Edge and Cloud Computing

The dramatic development of cloud and edge computing allows for better Quality of Service (QoS) in many scenarios by deploying services on cloud and edge servers. Microservice technology is also adopted in these scenarios to decompose complex business logic into many small independent services. Meanwhile, as microservice systems continue to grow, providing stable QoS in these systems becomes a challenge, and many different approaches have been proposed for stable QoS. However, the microservice systems used in the experiments of these work have problems such as the low number of services and a single type of service. Therefore, we developed the open-source benchmark microservice system RescureService with 20+ services, including database, front-end, business logic, data processing, and artificial intelligence services in the disaster relief scenario. Measuring tools are provided to measure the service properties to help researchers prepare experimental data, and the service properties pre-measured are also presented. Meanwhile, the fulfillment of benchmark requirements is detailed, and the results show that our RescureService meets the requirements of a benchmark system in research. Moreover, instructions are given to describe adopting our system in service computing as examples.

cs.DC↗

arXiv2022

Spike Calibration: Fast and Accurate Conversion of Spiking Neural Network for Object Detection and Segmentation

Spiking neural network (SNN) has been attached to great importance due to the properties of high biological plausibility and low energy consumption on neuromorphic hardware. As an efficient method to obtain deep SNN, the conversion method has exhibited high performance on various large-scale datasets. However, it typically suffers from severe performance degradation and high time delays. In particular, most of the previous work focuses on simple classification tasks while ignoring the precise approximation to ANN output. In this paper, we first theoretically analyze the conversion errors and derive the harmful effects of time-varying extremes on synaptic currents. We propose the Spike Calibration (SpiCalib) to eliminate the damage of discrete spikes to the output distribution and modify the LIPooling to allow conversion of the arbitrary MaxPooling layer losslessly. Moreover, Bayesian optimization for optimal normalization parameters is proposed to avoid empirical settings. The experimental results demonstrate the state-of-the-art performance on classification, object detection, and segmentation tasks. To the best of our knowledge, this is the first time to obtain SNN comparable to ANN on these tasks simultaneously. Moreover, we only need 1/50 inference time of the previous work on the detection task and can achieve the same performance under 0.492$\times$ energy consumption of ANN on the segmentation task.

cs.CV↗

arXiv2022

Degeneration of curves on some polarized toric surfaces

We address the following question: Given a polarized toric surface (S,L), and a general integral curve C of geometric genus g in the linear system |L|, do there exist degenerations of C in |L| to general integral curves of smaller geometric genera? We give an affirmative answer to this question for surfaces associated to h-transverse polygons, provided that the characteristic of the ground field is large enough. We give examples of surfaces in small characteristic, for which the answer to the question is negative. In case the answer is affirmative, we deduce that a general curve C as above is nodal. In characteristic 0, we use the result to show irreducibility of Severi varieties of a large class of polarized toric surfaces with h-transverse polygon.

math.AG↗

arXiv2021

Magic Pyramid: Accelerating Inference with Early Exiting and Token Pruning

Pre-training and then fine-tuning large language models is commonly used to achieve state-of-the-art performance in natural language processing (NLP) tasks. However, most pre-trained models suffer from low inference speed. Deploying such large models to applications with latency constraints is challenging. In this work, we focus on accelerating the inference via conditional computations. To achieve this, we propose a novel idea, Magic Pyramid (MP), to reduce both width-wise and depth-wise computation via token pruning and early exiting for Transformer-based models, particularly BERT. The former manages to save the computation via removing non-salient tokens, while the latter can fulfill the computation reduction by terminating the inference early before reaching the final layer, if the exiting condition is met. Our empirical studies demonstrate that compared to previous state of arts, MP is not only able to achieve a speed-adjustable inference but also to surpass token pruning and early exiting by reducing up to 70% giga floating point operations (GFLOPs) with less than 0.5% accuracy drop. Token pruning and early exiting express distinctive preferences to sequences with different lengths. However, MP is capable of achieving an average of 8.06x speedup on two popular text classification tasks, regardless of the sizes of the inputs.

cs.CL↗

arXiv2021

Degenerations of Grassmannians via lattice configurations

We study degenerations of Grassmannians constructed using convex lattice configurations in Bruhat-Tits buildings. Using techniques from quiver representations, we analyze their special fibers, which are explicitly described as quiver Grassmannians. For a class of lattice configurations, called the locally linearly independent configurations, we show that our construction coincide with Mustafin degenerations, thus generalizing a result of Faltings. In such cases, our analysis of special fibers also generalizes results of Cartwright et al. As an application, we prove a smoothing criterion for limit linear series on arbitrary reducible nodal curves.

math.AG↗

arXiv2021

Online Deployment Algorithms for Microservice Systems with Complex Dependencies

Cloud and edge computing have been widely adopted in many application scenarios. With the increasing demand of fast iteration and complexity of business logic, it is challenging to achieve rapid development and continuous delivery in such highly distributed cloud and edge computing environment. At present, microservice-based architecture has been the dominant deployment style, and a microservice system has to evolve agilely to offer stable Quality of Service (QoS) in the situation where user requirement changes frequently. Many research have been conducted to optimally re-deploy microservices to adapt to changing requirements. Nevertheless, complex dependencies between microservices and the existence of multiple instances of one single microservice in a microservice system have not been fully considered in existing works. This paper defines SPPMS, the Service Placement Problem in Microservice Systems that feature complex dependencies and multiple instances, as a Fractional Polynomial Problem (FPP) . Considering the high computation complexity of FPP, it is then transformed into a Quadratic Sum-of-Ratios Fractional Problem (QSRFP) which is further solved by the proposed greedy-based algorithms. Experiments demonstrate that our models and algorithms outperform existing approaches in both quality and computation speed.

cs.DC↗

arXiv2021

PAM: Understanding Product Images in Cross Product Category Attribute Extraction

Understanding product attributes plays an important role in improving online shopping experience for customers and serves as an integral part for constructing a product knowledge graph. Most existing methods focus on attribute extraction from text description or utilize visual information from product images such as shape and color. Compared to the inputs considered in prior works, a product image in fact contains more information, represented by a rich mixture of words and visual clues with a layout carefully designed to impress customers. This work proposes a more inclusive framework that fully utilizes these different modalities for attribute extraction. Inspired by recent works in visual question answering, we use a transformer based sequence to sequence model to fuse representations of product text, Optical Character Recognition (OCR) tokens and visual objects detected in the product image. The framework is further extended with the capability to extract attribute value across multiple product categories with a single model, by training the decoder to predict both product category and attribute value and conditioning its output on product category. The model provides a unified attribute extraction solution desirable at an e-commerce platform that offers numerous product categories with a diverse body of product attributes. We evaluated the model on two product attributes, one with many possible values and one with a small set of possible values, over 14 product categories and found the model could achieve 15% gain on the Recall and 10% gain on the F1 score compared to existing methods using text-only features.

cs.CV↗

arXiv2020

Brill-Noether Generality of Binary Curves

We show that the space of linear series of certain multi-degree (including the balanced ones) and rank $r$ on a general binary curve has the expected dimension if nonempty. This generalizes Theorem 24 of Caporaso's paper about binary curves from the case $r\leq 2$ to arbitrary rank, and shows that the space of Osserman-limit linear series on a general binary curve has the expected dimension, which was known for $r\leq 2$. In addition, we show that this space of linear series is still of expected dimension after imposing certain ramification conditions with respect to a sequence of increasing effective divisors supported on two general points lying on different components of the curve.

math.AG↗

arXiv2020

Secant planes of a general curve via degenerations

We study linear series on a general curve of genus g, whose images are exceptional with respect to their secant planes. Each such exceptional secant plane is algebraically encoded by an included linear series, whose number of base points computes the incidence degree of the corresponding secant plane. With enumerative applications in mind, we construct a moduli scheme of inclusions of limit linear series with base points over families of nodal curves of compact type, which we then use to compute combinatorial formulas for the number of secant-exceptional linear series when the spaces of linear series and of inclusions are finite.

math.AG↗

arXiv2020

AutoKnow: Self-Driving Knowledge Collection for Products of Thousands of Types

Can one build a knowledge graph (KG) for all products in the world? Knowledge graphs have firmly established themselves as valuable sources of information for search and question answering, and it is natural to wonder if a KG can contain information about products offered at online retail sites. There have been several successful examples of generic KGs, but organizing information about products poses many additional challenges, including sparsity and noise of structured data for products, complexity of the domain with millions of product types and thousands of attributes, heterogeneity across large number of categories, as well as large and constantly growing number of products. We describe AutoKnow, our automatic (self-driving) system that addresses these challenges. The system includes a suite of novel techniques for taxonomy construction, product property identification, knowledge extraction, anomaly detection, and synonym discovery. AutoKnow is (a) automatic, requiring little human intervention, (b) multi-scalable, scalable in multiple dimensions (many domains, many products, and many attributes), and (c) integrative, exploiting rich customer behavior logs. AutoKnow has been operational in collecting product knowledge for over 11K product types.

cs.AI↗

arXiv2019

Elliptically polarized laser-pumped $M_x$ magnetometer towards applications at room temperature

An atomic magnetometer operated with elliptically polarized light is investigated theoretically and experimentally. To explore the potential of this magnetometric configuration, the analytical form of the outgoing signal is derived. Parameters that significantly influence the performance are optimized, which lead to a sensitivity of 300 $\rm fT/\sqrt{Hz}$ at 45 $^{\circ}$C with a 2$\times$2$\times2$ cm uncoated Rb vapor cell. It is remarkable that a sensitivity of 690 $\rm fT/\sqrt{Hz}$ is achieved at room temperature of 24 $^{\circ}$C, which is improved by an order of magnitude compared with the conventional $M_x$ magnetometer under its own optimized condition. The elliptically polarized approach offers attractive features for developing compact, low-power magnetometers, which are available without heating the uncoated vapor cell.

physics.atom-ph↗

arXiv2019

Non-Local Context Encoder: Robust Biomedical Image Segmentation against Adversarial Attacks

Recent progress in biomedical image segmentation based on deep convolutional neural networks (CNNs) has drawn much attention. However, its vulnerability towards adversarial samples cannot be overlooked. This paper is the first one that discovers that all the CNN-based state-of-the-art biomedical image segmentation models are sensitive to adversarial perturbations. This limits the deployment of these methods in safety-critical biomedical fields. In this paper, we discover that global spatial dependencies and global contextual information in a biomedical image can be exploited to defend against adversarial attacks. To this end, non-local context encoder (NLCE) is proposed to model short- and long range spatial dependencies and encode global contexts for strengthening feature activations by channel-wise attention. The NLCE modules enhance the robustness and accuracy of the non-local context encoding network (NLCEN), which learns robust enhanced pyramid feature representations with NLCE modules, and then integrates the information across different levels. Experiments on both lung and skin lesion segmentation datasets have demonstrated that NLCEN outperforms any other state-of-the-art biomedical image segmentation methods against adversarial attacks. In addition, NLCE modules can be applied to improve the robustness of other CNN-based biomedical image segmentation methods.

cs.CV↗

arXiv2018

A Generalization of Lifting Non-proper Tropical Intersections

Let X and X' be closed subschemes of an algebraic torus T over a non-archimedean field. We prove the rational equivalence as tropical cycles, in the sense of Henning Meyer's graduate thesis, between the tropicalization of the intersection product of X and X' and the stable intersection of trop(X) and trop(X'), when restricted to (the inverse image under the tropicalization map of) a connected component C of the intersection of trop(X) and trop(X'). This requires possibly passing to a (partial) compactification of T with respect to a suitable fan. We define the compactified stable intersection in a toric tropical variety, and check that this definition is compatible with the intersection product in loc.cit.. As a result we get a numerical equivalence (after a compactification and restricting to C) between the intersection product of X and X' and the stable intersection of trop(X) and trop(X') via the compactified stable intersection. In particular, when X and X' have complementary codimensions, this equivalence generalizes the work of Osserman and Rabinoff, in the sense that the intersection of X and X' is allowed to be of positive dimension. Moreover, if the intersection of the closures of X and X' has finitely many points which tropicalize to the closure of C, we prove a similar equation as in Theorem 6.4 of the paper of Osserman and Rabinoff when the ambient space is a reduced closed subscheme of T (instead of T itself).

math.AG↗

arXiv2018

Non-locally Enhanced Encoder-Decoder Network for Single Image De-raining

Single image rain streaks removal has recently witnessed substantial progress due to the development of deep convolutional neural networks. However, existing deep learning based methods either focus on the entrance and exit of the network by decomposing the input image into high and low frequency information and employing residual learning to reduce the mapping range, or focus on the introduction of cascaded learning scheme to decompose the task of rain streaks removal into multi-stages. These methods treat the convolutional neural network as an encapsulated end-to-end mapping module without deepening into the rationality and superiority of neural network design. In this paper, we delve into an effective end-to-end neural network structure for stronger feature expression and spatial correlation learning. Specifically, we propose a non-locally enhanced encoder-decoder network framework, which consists of a pooling indices embedded encoder-decoder network to efficiently learn increasingly abstract feature representation for more accurate rain streaks modeling while perfectly preserving the image detail. The proposed encoder-decoder framework is composed of a series of non-locally enhanced dense blocks that are designed to not only fully exploit hierarchical features from all the convolutional layers but also well capture the long-distance dependencies and structural information. Extensive experiments on synthetic and real datasets demonstrate that the proposed method can effectively remove rain-streaks on rainy image of various densities while well preserving the image details, which achieves significant improvements over the recent state-of-the-art methods.

cs.CV↗