arXiv subjects

Shuai Zhao

Publications and source records attributed to Shuai Zhao.

At least 163 records · Page 9Linked to original sources

arXiv2019

DBP: Discrimination Based Block-Level Pruning for Deep Model Acceleration

Neural network pruning is one of the most popular methods of accelerating the inference of deep convolutional neural networks (CNNs). The dominant pruning methods, filter-level pruning methods, evaluate their performance through the reduction ratio of computations and deem that a higher reduction ratio of computations is equivalent to a higher acceleration ratio in terms of inference time. However, we argue that they are not equivalent if parallel computing is considered. Given that filter-level pruning only prunes filters in layers and computations in a layer usually run in parallel, most computations reduced by filter-level pruning usually run in parallel with the un-reduced ones. Thus, the acceleration ratio of filter-level pruning is limited. To get a higher acceleration ratio, it is better to prune redundant layers because computations of different layers cannot run in parallel. In this paper, we propose our Discrimination based Block-level Pruning method (DBP). Specifically, DBP takes a sequence of consecutive layers (e.g., Conv-BN-ReLu) as a block and removes redundant blocks according to the discrimination of their output features. As a result, DBP achieves a considerable acceleration ratio by reducing the depth of CNNs. Extensive experiments show that DBP has surpassed state-of-the-art filter-level pruning methods in both accuracy and acceleration ratio. Our code will be made available soon.

cs.CV↗

arXiv2019

Region Mutual Information Loss for Semantic Segmentation

Semantic segmentation is a fundamental problem in computer vision. It is considered as a pixel-wise classification problem in practice, and most segmentation models use a pixel-wise loss as their optimization riterion. However, the pixel-wise loss ignores the dependencies between pixels in an image. Several ways to exploit the relationship between pixels have been investigated, \eg, conditional random fields (CRF) and pixel affinity based methods. Nevertheless, these methods usually require additional model branches, large extra memories, or more inference time. In this paper, we develop a region mutual information (RMI) loss to model the dependencies among pixels more simply and efficiently. In contrast to the pixel-wise loss which treats the pixels as independent samples, RMI uses one pixel and its neighbour pixels to represent this pixel. Then for each pixel in an image, we get a multi-dimensional point that encodes the relationship between pixels, and the image is cast into a multi-dimensional distribution of these high-dimensional points. The prediction and ground truth thus can achieve high order consistency through maximizing the mutual information (MI) between their multi-dimensional distributions. Moreover, as the actual value of the MI is hard to calculate, we derive a lower bound of the MI and maximize the lower bound to maximize the real value of the MI. RMI only requires a few extra computational resources in the training stage, and there is no overhead during testing. Experimental results demonstrate that RMI can achieve substantial and consistent improvements in performance on PASCAL VOC 2012 and CamVid datasets. The code is available at https://github.com/ZJULearning/RMI.

cs.CV↗

arXiv2019

Correlation Maximized Structural Similarity Loss for Semantic Segmentation

Most semantic segmentation models treat semantic segmentation as a pixel-wise classification task and use a pixel-wise classification error as their optimization criterions. However, the pixel-wise error ignores the strong dependencies among the pixels in an image, which limits the performance of the model. Several ways to incorporate the structure information of the objects have been investigated, \eg, conditional random fields (CRF), image structure priors based methods, and generative adversarial network (GAN). Nevertheless, these methods usually require extra model branches or additional memories, and some of them show limited improvements. In contrast, we propose a simple yet effective structural similarity loss (SSL) to encode the structure information of the objects, which only requires a few additional computational resources in the training phase. Inspired by the widely-used structural similarity (SSIM) index in image quality assessment, we use the linear correlation between two images to quantify their structural similarity. And the goal of the proposed SSL is to pay more attention to the positions, whose associated predictions lead to a low degree of linear correlation between two corresponding regions in the ground truth map and the predicted map. Thus the model can achieve a strong structural similarity between the two maps through minimizing the SSL over the whole map. The experimental results demonstrate that our method can achieve substantial and consistent improvements in performance on the PASCAL VOC 2012 and Cityscapes datasets. The code will be released soon.

cs.CV↗

arXiv2019

Integrated Process Planning and Scheduling in Commercial Smart Kitchens

This paper describes the possibility of applying a generic, cloud-based Optimisation as a Service facility to food cooking planning and scheduling in a commercial kitchen. We propose a chromosome encoding and customisation of the classic MOEA/D multi-objective genetic algorithm. The applicability of the proposed approach is evaluated experimentally for two scenarios different with respect to the number of cooking appliances and the amount of the ordered food. The proposed system managed to determine the trade-offs between cooking time, energy dissipation and food quality.

cs.DC↗

arXiv2019

An XML-based Factory Description Language for Smart Manufacturing Plants in Industry 4.0

Industry 4.0 revolution concerns the digital transformation of manufacturing and promises to answer the ever-increasing demand of product customisation and manufacturing flexibility while incurring low costs. To perform the required factory reconfiguration, a computationally demanding optimisation process has to be executed to find favourable solutions in a relatively short time. While previous research focused on planning and scheduling of smart factories based on cloud-based optimisation, little attention has been paid to effective approaches to describe the targeted factory, the required products and the production processes. However, these matters are fundamental for the optimisation engine to be correctly and efficiently performed. This paper presents an XML-based factory modelling language to effectively describe the above data for a given factory and commodity order and to provide a convenient interface for altering the input information. Finally, two real-world manufacturing plants are provided to illustrate the feasibility of the proposed description language.

cs.DC↗

arXiv2019

ACE-Net: Biomedical Image Segmentation with Augmented Contracting and Expansive Paths

Nowadays U-net-like FCNs predominate various biomedical image segmentation applications and attain promising performance, largely due to their elegant architectures, e.g., symmetric contracting and expansive paths as well as lateral skip-connections. It remains a research direction to devise novel architectures to further benefit the segmentation. In this paper, we develop an ACE-net that aims to enhance the feature representation and utilization by augmenting the contracting and expansive paths. In particular, we augment the paths by the recently proposed advanced techniques including ASPP, dense connection and deep supervision mechanisms, and novel connections such as directly connecting the raw image to the expansive side. With these augmentations, ACE-net can utilize features from multiple sources, scales and reception fields to segment while still maintains a relative simple architecture. Experiments on two typical biomedical segmentation tasks validate its effectiveness, where highly competitive results are obtained in both tasks while ACE-net still runs fast at inference.

cs.CV↗

arXiv2019

Matching Quasi Generalized Parton Distributions in the RI/MOM scheme

Within the framework of large momentum effective theory (LaMET), genenaralized parton distributions (GPDs) can be extracted from lattice calculations of quasi-GPDs through a perturbative matching relation, up to power corrections that are suppressed by the hadron momentum. In this paper, we focus on isovector quark GPDs, including the unpolarized, longitudinally and transversely polarized cases, and present the one-loop matching that connects the quasi-GPDs renormalized in a regularization-independent momentum subtraction (RI/MOM) scheme to the GPDs in MS scheme. We find that the matching coefficient is independent of the momentum transfer squared. As a consequence, the matching for the quasi-GPD with zero skewness is the same as that for the quasi-PDF. Our results provide a crucial input for the determination of quark GPDs from lattice QCD using LaMET.

hep-ph↗

arXiv2019

Improving Semantic Segmentation via Dilated Affinity

Introducing explicit constraints on the structural predictions has been an effective way to improve the performance of semantic segmentation models. Existing methods are mainly based on insufficient hand-crafted rules that only partially capture the image structure, and some methods can also suffer from the efficiency issue. As a result, most of the state-of-the-art fully convolutional networks did not adopt these techniques. In this work, we propose a simple, fast yet effective method that exploits structural information through direct supervision with minor additional expense. To be specific, our method explicitly requires the network to predict semantic segmentation as well as dilated affinity, which is a sparse version of pair-wise pixel affinity. The capability of telling the relationships between pixels are directly built into the model and enhance the quality of segmentation in two stages. 1) Joint training with dilated affinity can provide robust feature representations and thus lead to finer segmentation results. 2) The extra output of affinity information can be further utilized to refine the original segmentation with a fast propagation process. Consistent improvements are observed on various benchmark datasets when applying our framework to the existing state-of-the-art model. Codes will be released soon.

cs.CV↗

arXiv2019

Proceedings of the International Workshop on Reconfigurable and Communication-centric Cyber-Physical Systems (ReCoCyPS 2019)

This volume represents the proceedings of the International Workshop on Reconfigurable and Communication-centric Cyber-Physical Systems (ReCoCyPS 2019), co-located with the 14th International Symposium on Reconfigurable Communication-centric Systems-on-Chip (ReCoSoC 2019) on July 2-3 2019, York, United Kingdom. The workshop is organised by the EU-funded SAFIRE project consortium.

eess.SP↗

arXiv2019

Model Based System Assurance Using the Structured Assurance Case Metamodel

Assurance cases are used to demonstrate confidence in system properties of interest (e.g. safety and/or security). A number of system assurance approaches are adopted by industries in the safety-critical domain. However, the task of constructing assurance cases remains a manual, trivial and informal process. The Structured Assurance Case Metamodel (SACM) is a standard specified by the Object Management Group (OMG). SACM provides a richer set of features than existing system assurance languages/approaches. SACM provides a foundation for model-based system assurance, which has great potentials in growing technology domains such as Open Adaptive Systems. However, the intended usage of SACM has not been sufficiently explained. In addition, there has been no support to interoperate between existing assurance case (models) and SACM models. In this article, we explain the intended usage of SACM based on our involvement in the OMG specification process of SACM. In addition, to promote a model-based approach, we provide SACM compliant metamodels for existing system assurance approaches (the Goal Structuring Notation and Claims-Arguments-Evidence), and the transformations from these models to SACM. We also briefly discuss the tool support for model-based system assurance which helps practitioners to make the transition from existing system assurance approaches to model-based system assurance using SACM.

cs.SE↗

arXiv2019

Accessing gluon parton distributions in large momentum effective theory

Gluon parton distribution functions (PDFs) in the proton can be calculated directly on Euclidean lattices using large momentum effective theory (LaMET). To realize this goal, one has to find renormalized gluon quasi-PDFs in which power divergences and operator mixing are thoroughly understood. For the unpolarized distribution, we identify four independent quasi-PDF correlators that can be multiplicatively renormalized on the lattice. Similarly, the helicity distribution can be derived from three independent multiplicatively renormalizable quasi-PDFs. We provide a LaMET factorization formula for these renormalized quasi-PDFs from which one can extract the gluon PDFs.

hep-ph↗

arXiv2018

Matching the Quasi Meson Distribution Amplitude in RI/MOM scheme

The $x$-dependence of light-cone distribution amplitude (LCDA) can be directly calculated from a quasi distribution amplitude (DA) in lattice QCD within the framework of large-momentum effective theory (LaMET). In this paper, we study the one-loop renormalization of the quasi-DA in the regularization-independent momentum subtraction (RI/MOM) scheme. The renormalization factor for the quasi parton distribution function can be used to renormalize the quasi-DA provided that they are implemented on lattice and in perturbation theory in the same manner. We derive the one-loop matching coefficient that matches quasi-DA in the RI/MOM scheme onto LCDA in the $\overline{\rm MS}$ scheme. Our result provides the crucial step to extract the LCDAs from lattice matrix elements of quasi-DAs.

hep-ph↗

arXiv2018

Packaging and Sharing Machine Learning Models via the Acumos AI Open Platform

Applying Machine Learning (ML) to business applications for automation usually faces difficulties when integrating diverse ML dependencies and services, mainly because of the lack of a common ML framework. In most cases, the ML models are developed for applications which are targeted for specific business domain use cases, leading to duplicated effort, and making reuse impossible. This paper presents Acumos, an open platform capable of packaging ML models into portable containerized microservices which can be easily shared via the platform's catalog, and can be integrated into various business applications. We present a case study of packaging sentiment analysis and classification ML models via the Acumos platform, permitting easy sharing with others. We demonstrate that the Acumos platform reduces the technical burden on application developers when applying machine learning models to their business applications. Furthermore, the platform allows the reuse of readily available ML microservices in various business domains.

cs.AI↗

arXiv2018

On the power divergence in quasi gluon distribution function

Recent perturbative calculation of quasi gluon distribution function at one-loop level shows the existence of extra linear ultraviolet divergences in the cut-off scheme. We employ the auxiliary field approach, and study the renormalization of gluon operators. The non-local gluon operator can mix with new operators under renormalization, and the linear divergences in quasi distribution function can be into the newly introduced operators. After including the mixing, we find the improved quasi gluon distribution functions contain only logarithmic divergences, and thus can be used to extract the gluon distribution in large momentum effective theory.

hep-ph↗

arXiv2018

Light-cone distribution amplitudes of vector meson in large momentum effective theory

We investigate the leading twist light-cone distribution amplitudes (LCDAs) of vector meson in the framework of large momentum effective theory. We derive the matching equation for the LCDAs and quasi distribution amplitudes. The matching coefficients are determined to one loop accuracy, both in the ultraviolet cut-off and dimensional regularization schemes. This calculation provides the possibility of studying the full $x$ behavior of LCDAs and extracting LCDAs of vector mesons from lattice simulations.

hep-ph↗

arXiv2018

Gluon quasidistribution function at one loop

We study the unpolarized gluon quasidistribution function in the nucleon at one loop level in the large momentum effective theory. For the quark quasidistribution, power law ultraviolet divergences arise in the cut-off scheme and an important observation is that they all are subjected to Wilson lines. However for the gluon quasidistribution function, we first point out that the linear ultraviolet divergences also exist in the real diagram which is not connected to any Wilson line. We then study the one loop corrections to parton distribution functions in both cut-off scheme and dimensional regularization to deal with the ultraviolet divergences. In addition to the ordinary quark and gluon distributions, we also include the quark to gluon and gluon to quark splitting diagrams. The complete one-loop matching factors between the quasi and light cone parton distribution functions are presented in the cut-off scheme. We derive the $P^z$ evolution equation for quasi parton distribution functions, and find that the $P^z$ evolution kernels are identical to the DGLAP evolution kernels.

hep-ph↗

arXiv2017

Implications of the $R_K$ and $R_{K^*}$ anomalies

We discuss the implications of the recently reported $R_K$ and $R_{K^*}$ anomalies, the lepton flavor non-universality in the $B\to K\ell^+\ell^-$ and $B\to K^*\ell^+\ell^-$. Using two sets of hadronic inputs of form factors, we perform a fit of the new physics to the $R_K$ and $R_{K^*}$ data, and significant new physics contributions are found. We suggest to study the lepton flavor universality in a number of related rare $B, B_s, B_c$ and $Λ_b$ decay channels, and in particular we give the predictions for the $μ$-to-$e$ ratios of decay widths with different polarizations of the final state particles, and of the $b\to d\ell^+\ell^-$ processes which are presumably more sensitive to the structure of the underlying new physics. With the new physics contributions embedded in Wilson coefficients, we present theoretical predictions for lepton flavor non-universality in these processes.

hep-ph↗

arXiv2017

Chiral dynamics, S-wave contributions and angular analysis in $D\to ππ\ell\barν$

We present a theoretical analysis of the $D^-\to π^+π^- \ell\barν$ and $\bar D^0\to π^+π^0 \ell\barν$ decays. We construct a general angular distribution which can include arbitrary partial waves of $ππ$. Retaining the S-wave and P-wave contributions we study the branching ratios, forward--backward asymmetries and a few other observables. The P-wave contribution is dominated by $ρ^0$ resonance, and the S-wave contribution is analyzed using the unitarized chiral perturbation theory. The obtained branching fraction for $D\to ρ\ellν$, at the order $10^{-3}$, is consistent with the available experimental data. The S-wave contribution has a branching ratio at the order of $10^{-4}$, and this prediction can be tested by experiments like BESIII and LHCb. Future measurements can also be used to examine the $π$--$π$ scattering phase shift.

hep-ph↗