Direct training of hardware-friendly weight binarized spiking neural network with surrogate gradient learning towards spatio-temporal event-based dynamic data recognition

Qiao, Guanchao; Ning, Ning; Zuo, Yuan; Hu, Shaogang; Yu, Qin; Liu, Y.

doi:10.1016/j.neucom.2021.06.070

Cited by 15 publications

(2 citation statements)

References 29 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Lin [22] defined two improvements in terms of network cohesion and network edge importance using weighted shortest distance to propose an improved importance evaluation method. The network weighting information is used in neural networks for applications such as efficient image classification and hardware dynamic information processing of neuromorphic forms [23,24].…”

Section: Introductionmentioning

confidence: 99%

Identification of node rankings in temporal networks based on multilayer topological overlap coefficients

Niu

Xu³

et al. 2023

Phys. Scr.

View full text Add to dashboard Cite

Identifying node ranking in complex networks over time is a crucial research topic. The topology relationship of general network nodes reflects their importance in the network. The node ranking evolution within the temporal layers depends not only on the current layer's topology relationship but also on the nodes' interaction relationships as they evolve. In this study, we propose a method called the multilayer topological overlap coefficient-based supra-adjacency matrix to identify node rankings. To account for the node evolution process, we analyze and establish the node ranking matrix structure of unweighted and weighted temporal networks in the temporal network. We also analyze the sequence multilayer node topological overlap structure throughout the whole-time layer. The experimental results demonstrate that the topological overlap coefficient unweighted supra-adjacency matrix of multilayer nodes performs up to 15.00% and 25.80% better than the two supra-adjacency matrix metrics under three different metrics. Moreover, the topological overlap coefficient weighted supra-adjacency matrix of multilayer nodes outperforms the SAM metrics by up to 70.20%.

show abstract

Section: Introductionmentioning

confidence: 99%

Identification of node rankings in temporal networks based on multilayer topological overlap coefficients

Niu

Xu³

et al. 2023

Phys. Scr.

View full text Add to dashboard Cite

show abstract

“…Recently, a gradient surrogate method was proposed to train deeper SNNs with multiple hidden layers [25][26][27][28][29][30][31]. With this algorithm, the spiking nonlinearity derivation was replaced by the derivation of a continuously differentiable function.…”

Section: Introductionmentioning

confidence: 99%

Spiking VGG7: Deep Convolutional Spiking Neural Network with Direct Training for Object Recognition

et al. 2022

View full text Add to dashboard Cite

We propose a deep convolutional spiking neural network (DCSNN) with direct training to classify concrete bridge damage in a real engineering environment. The leaky-integrate-and-fire (LIF) neuron model is employed in our DCSNN that is similar to VGG. Poisson encoding and convolution encoding strategies are considered. The gradient surrogate method is introduced to realize the supervised training for the DCSNN. In addition, we have examined the effect of observation time step on the network performance. The testing performance for two different spike encoding strategies are compared. The results show that the DCSNN using gradient surrogate method can achieve a performance of 97.83%, which is comparable to traditional CNN. We also present a comparison with STDP-based unsupervised learning and a converted algorithm, and the proposed DCSNN is proved to have the best performance. To demonstrate the generalization performance of the model, we also use a public dataset for comparison. This work paves the way for the practical engineering applications of the deep SNNs.

show abstract

Efficient Training of Spiking Neural Networks with Multi-parallel Implicit Stream Architecture

Cao,

Li,

Wang

et al. 2024

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Direct training of hardware-friendly weight binarized spiking neural network with surrogate gradient learning towards spatio-temporal event-based dynamic data recognition

Cited by 15 publications

References 29 publications

Identification of node rankings in temporal networks based on multilayer topological overlap coefficients

Identification of node rankings in temporal networks based on multilayer topological overlap coefficients

Spiking VGG7: Deep Convolutional Spiking Neural Network with Direct Training for Object Recognition

Efficient Training of Spiking Neural Networks with Multi-parallel Implicit Stream Architecture

Contact Info

Product

Resources

About