Dynamic Fixed-Point Design of Neuromorphic Computing Systems

Yongshin KANG  Jaeyong CHUNG  

IEICE TRANSACTIONS on Electronics   Vol.E101-C   No.10   pp.840-844
Publication Date: 2018/10/01
Online ISSN: 1745-1353
DOI: 10.1587/transele.E101.C.840
Type of Manuscript: BRIEF PAPER
Category: Microwaves, Millimeter-Waves
neural network,  machine learning,  VLSI,  

Full Text: PDF(348.7KB)
>>Buy this Article

Practical deep neural networks have a number of weight parameters, and the dynamic fixed-point formats have been used to represent them efficiently. The dynamic fixed-point representations share an scaling factor among a group of numbers, and the weights in a layer have been formed into such a group. In this paper, we first explore a design space for dynamic fixed-point neuromorphic computing systems and show that it is indispensable to have a small group size in neuromorphic architectures, because it is appropriate to group the weights associated with a neuron into a group. We then presents a dynamic fixed-point representation designed for neuromorphic computing systems. Our experimental results show that the proposed representation reduces the required weight bitwidth by about 4 bits compared to the conventional fixed-point format.