Abstract
This chapter summarizes applications of stochastic computing for braininspired computing, which we refer to as Brainware Large-Scale Integration (BLSI). Stochastic computing exploits random bit streams, realizing area-efficient hardware for complicated functions such as multiplication and tanh, as compared with more traditional binary approaches. Using stochastic computing, we have implemented hardware for several physiological models of the primary visual cortex of brains, where these models require such complicated functions. In addition, a deep neural network using stochastic computing has been designed for area/energy-efficient hardware. In order to design BLSIs, we have introduced extended arithmetic functions, such as circular functions. As a design example, our BLSIs are implemented using Taiwan Semiconductor Manufacturing Company (TSMC) 65-nm Complementary Metal Oxide Semiconductor (CMOS) and discussed with traditional fixed-point implementations in terms of hardware performance and computation accuracy.
Original language | English |
---|---|
Title of host publication | Stochastic Computing |
Subtitle of host publication | Techniques and Applications |
Publisher | Springer International Publishing |
Pages | 185-199 |
Number of pages | 15 |
ISBN (Electronic) | 9783030037307 |
ISBN (Print) | 9783030037291 |
DOIs | |
Publication status | Published - 2019 Feb 18 |
Keywords
- Deep neural networks
- Integrated circuits
- Neuromorphic computing
ASJC Scopus subject areas
- Engineering(all)
- Computer Science(all)
- Mathematics(all)