Nguyen, V.-T. and Luong, T.-K. and Le Duc, H. and Hoang, V.-P. (2018) An efficient hardware implementation of activation functions using stochastic computing for deep neural networks. In: 12th IEEE International Symposium on Embedded Multicore/Many-Core Systems-on-Chip, MCSoC 2018, 12 September 2018 through 14 September 2018.
An efficient hardware implementation of activation functions using stochastic computing for deep neural networks..pdf
Download (444kB) | Preview
Abstract
In this paper, we present a new approximation method for non-linear activation functions including tanh and sigmoid functions using stochastic computing (SC) logic based on the piecewise-linear approximation (PWL) for the full range of [-1, 1]. SC implementations with PWL approximation expansions for non-linear functions are based on a 90nm CMOS process. The implementation results shown that the proposed SC circuits can provide better performance compared with the previous methods such as the well-known Maclaurin expansions based, Bernstein polynomial based and finite-state-machine (FSM) based implementations. The implementation results are also presented and discussed. © 2018 IEEE.
Item Type: | Conference or Workshop Item (Paper) |
---|---|
Divisions: | Faculties > Faculty of Radio-Electronic Engineering |
Identification Number: | 10.1109/MCSoC2018.2018.00045 |
Uncontrolled Keywords: | Chemical activation; CMOS integrated circuits; Computation theory; Embedded systems; Functions; Hardware; Piecewise linear techniques; Stochastic systems; Activation functions; Approximation methods; Bernstein polynomial; Hardware implementations; Maclaurin expansion; Nonlinear activation functions; Piecewise linear approximations; Stochastic computing; Deep neural networks |
Additional Information: | Conference code: 142785. Language of original document: English. |
URI: | http://eprints.lqdtu.edu.vn/id/eprint/9503 |