Heterogeneous Activation Function Extraction for Training and Optimization of SNN Systems

None, None; None, None; None, None

Heterogeneous Activation Function Extraction for Training and Optimization of SNN Systems

Conference Paper (2019)

Author(s)

Amir Zjajo (TU Delft - Electrical Engineering, Mathematics and Computer Science)

Sumeet Kumar (TU Delft - Electrical Engineering, Mathematics and Computer Science)

Rene Van Leuken (TU Delft - Electrical Engineering, Mathematics and Computer Science)

Research Group

Signal Processing Systems

Training Mismatch Extraction Deep neural network Spiking neural network Activation functions

DOI related publication

https://doi.org/10.1109/AICAS.2019.8771619 Final published version

To reference this document use

https://resolver.tudelft.nl/uuid:7bf40973-7fe1-4109-a414-9917c2861814

More Info

expand_more

Publication Year

2019

Language

English

Research Group

Signal Processing Systems

Article number

8771619

Pages (from-to)

244-245

ISBN (print)

978-1-5386-7885-5

ISBN (electronic)

978-1-5386-7884-8

Event

1st IEEE International Conference on Artificial Intelligence Circuits and Systems, AICAS 2019 (2019-03-18 - 2019-03-20), Hsinchu, Taiwan

Downloads counter

186

Abstract

Energy-efficiency and computation capability characteristics of analog/mixed-signal spiking neural networks offer capable platform for implementation of cognitive tasks on resource-limited embedded platforms. However, inherent mismatch in analog devices severely influence accuracy and reliability of the computing system. In this paper, we devise efficient algorithm for extracting of heterogeneous activation functions of analog hardware neurons as a set of constraints in an off-line training and optimization process, and examine how compensation of the mismatch effects influence synchronicity and information processing capabilities of the system.