A stochastic learning algorithm for neuromemristive systems

Author

Merkel, Cory ; Kudithipudi, Dhireesha

Author_Institution

Dept. of Comput. Eng., Rochester Inst. of Technol., Rochester, NY, USA

fYear

2014

fDate

2-5 Sept. 2014

Firstpage

359

Lastpage

364

Abstract

In this paper, we present a stochastic learning algorithm for neuromemristive systems. Existing algorithms are based on gradient descent techniques, which require analog multiplications. The proposed algorithm removes the necessity for an analog multiplier by transforming each variable into a random Bernoulli-distributed value. Arithmetic operations on such values are easily implemented using digital circuits, reducing the area cost of the implementation. We tested the proposed algorithm and compared with the least-mean-squares algorithm on both linear and logistic regression problems. Results indicate that the proposed algorithm is able to achieve similar accuracy with up to ≈3.5× less area.

Keywords

gradient methods; learning (artificial intelligence); least mean squares methods; memristors; neural chips; regression analysis; stochastic processes; Bernoulli-distributed value; analog multiplications; analog multiplier; arithmetic operations; digital circuits; gradient descent techniques; least-mean-squares algorithm; linear regression problems; logistic regression problems; neuromemristive systems; stochastic learning algorithm; Algorithm design and analysis; Hardware; Least squares approximations; Linear regression; Random variables; Training; Transistors;

fLanguage

English

Publisher

ieee

Conference_Titel

System-on-Chip Conference (SOCC), 2014 27th IEEE International

Conference_Location

Las Vegas, NV

Type

conf

DOI

10.1109/SOCC.2014.6948954

Filename

6948954