FloatSD: A New Weight Representation and Associated Update Method for Efficient Convolutional Neural Network Training
Sign Up to like & getrecommendations! Published in 2019 at "IEEE Journal on Emerging and Selected Topics in Circuits and Systems"
DOI: 10.1109/jetcas.2019.2911999
Abstract: In this paper, we propose a floating-point signed digit (FloatSD) format for convolutional neural network (CNN) weight representation and its update method during CNN training. The number of non-zero digits in a weight can be… read more here.
Keywords: floating point; method; weight representation; point ... See more keywords