Enhancing synchronizability by weight randomization on regular networks

Li, D. Q.; Li, M. H.; Wu, Jinshan; Di, Zengru; Yang, Fan

doi:10.1140/epjb/e2007-00189-0

Cited by 10 publications

(6 citation statements)

References 20 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Therefore, the improved mean-field equation is proposed based on the analysis of the root cause. In addition, the results of the comparison between the simulations and theoretical verification demonstrate that the modified equation is more accordant with simulations than the unimproved one on different kinds of networks [31,32]. The contents of this paper are divided into the following sections.…”

Section: Introductionmentioning

confidence: 88%

Comparison of two mean-field based theoretical analysis methods for SIS model

Zhang

Yang

2017

Chaos, Solitons & Fractals

View full text Add to dashboard Cite

Epidemic spreading has been intensively studied in SIS epidemic model. Although the mean-field theory of SIS model has been widely used in the research, there is a lack of comparative results between different theoretical calculations, and the differences between them should be systematically explained. In this paper, we have compared different theoretical solutions for mean-field theory and explained the underlying reason. We first describe the differences between different equations for mean-field theory in different networks. The results show that the difference between mean-field reaction equations is due to the different probability consideration for the infection process. This finding will help us to design better theoretical solutions for epidemic models.

show abstract

Section: Introductionmentioning

confidence: 88%

Comparison of two mean-field based theoretical analysis methods for SIS model

Zhang

Yang

2017

Chaos, Solitons & Fractals

View full text Add to dashboard Cite

show abstract

“…Although there are different implementations that leverage quantization techniques by transforming the weights into power of two values [10], or more aggressive ternary [11], or even binary values [12,13,14] and multiplications are hence downgraded to pure shifting or adding operations, these schemes suffer from severe accuracy loss at large datasets. High precision MACs are sometimes necessary and how to effectively optimize zero values and zero-valued bits simultaneously remains as a critical problem in DNN accelerator design.…”

Section: A Ineffectual Computationsmentioning

confidence: 99%

Architecting Effectual Computation for Machine Learning Accelerators

Zhang

Han

et al. 2020

IEEE Trans. Comput.-Aided Des. Integr. Circuits Syst.

View full text Add to dashboard Cite

Inference efficiency is the predominant consideration in designing deep learning accelerators. Previous work mainly focuses on skipping zero values to deal with remarkable ineffectual computation, while zero bits in non-zero values, as another major source of ineffectual computation, is often ignored. The reason lies on the difficulty of extracting essential bits during operating multiply-and-accumulate (MAC) in the processing element. Based on the fact that zero bits occupy as high as 68.9% fraction in the overall weights of modern deep convolutional neural network models, this paper firstly proposes a weight kneading technique that could eliminate ineffectual computation caused by either zero value weights or zero bits in non-zero weights, simultaneously. Besides, a split-and-accumulate (SAC) computing pattern in replacement of conventional MAC, as well as the corresponding hardware accelerator design called Tetris are proposed to support weight kneading at the hardware level. Experimental results prove that Tetris could speed up inference up to 1.50x, and improve power efficiency up to 5.33x compared with the state-of-the-art baselines.

show abstract

“…The small-world model [ 9 ] proposed by Watts and Strogatz captures the features of high clustering and small average path length, which have been widely observed in real-world networks. These small-world features [ 21 , 22 ] have been found to have a significant impact on the dynamics in interconnected networks [ 23 , 24 ]. The small-world model can be constructed from a regular lattice.…”

Section: Introductionmentioning

confidence: 99%

Epidemics in Interconnected Small-World Networks

Qin

et al. 2015

PLoS ONE

View full text Add to dashboard Cite

Networks can be used to describe the interconnections among individuals, which play an important role in the spread of disease. Although the small-world effect has been found to have a significant impact on epidemics in single networks, the small-world effect on epidemics in interconnected networks has rarely been considered. Here, we study the susceptible-infected-susceptible (SIS) model of epidemic spreading in a system comprising two interconnected small-world networks. We find that the epidemic threshold in such networks decreases when the rewiring probability of the component small-world networks increases. When the infection rate is low, the rewiring probability affects the global steady-state infection density, whereas when the infection rate is high, the infection density is insensitive to the rewiring probability. Moreover, epidemics in interconnected small-world networks are found to spread at different velocities that depend on the rewiring probability.

show abstract

Enhancing synchronizability by weight randomization on regular networks

Cited by 10 publications

References 20 publications

Comparison of two mean-field based theoretical analysis methods for SIS model

Comparison of two mean-field based theoretical analysis methods for SIS model

Architecting Effectual Computation for Machine Learning Accelerators

Epidemics in Interconnected Small-World Networks

Contact Info

Product

Resources

About