Spike-Train Level Unsupervised Learning Algorithm for Deep Spiking Belief Networks

Lin, Xianghong; Du, Pangao

doi:10.1007/978-3-030-61616-8_51

Cited by 1 publication

(1 citation statement)

References 20 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The dynamic threshold process in Section 2.2.1 is similar to how existing methods incorporate membrane potential values in the error signal (Gutig, 2016;Xiao et al, 2019;Yu et al, 2019;Li and Yu, 2020), but without strict precision requirements. In principle, the proposed error function may be extended to train deep or recurrent architectures using techniques which propagate error gradients based on the Widrow-Hoff window, such as Wang et al (2016); Lin and Shi (2018); Lin and Du (2020). The missing component is how to correctly incorporate the proposed adaptive 'learning rate' variables in such methods, which we leave to future work.…”

Section: Discussionmentioning

confidence: 99%

Linear Constraints Learning for Spiking Neurons

Nguyêݱn

Chu²

2021

Preprint

View full text Add to dashboard Cite

Encoding information with precise spike timings using spike-coded neurons has been shown to be more computationally powerful than rate-coded approaches. However, most existing supervised learning algorithms for spiking neurons are complicated and offer poor time complexity. To address these limitations, we propose a supervised multi-spike learning algorithm which reduces the required number of training iterations. We achieve this by formulating a large number of weight updates as a linear constraint satisfaction problem, which can be solved efficiently. Experimental results show this method offers better efficiency compared to existing algorithms on the MNIST dataset. Additionally, we provide experimental results on the classification capacity of the LIF neuron model, relative to several parameters of the system.

show abstract