HAW: Hardware-Aware Point Selection for Efficient Winograd Convolution
Sign Up to like & getrecommendations! Published in 2023 at "IEEE Signal Processing Letters"
DOI: 10.1109/lsp.2023.3258863
Abstract: Winograd's minimal filtering algorithm effectively reduces the multiplication arithmetic complexity of Convolutional Neural Networks. However, Winograd convolutions in current implementations are limited to small feature tiles for two reasons: the numerical error and the overhead… read more here.
Keywords: aware point; hardware aware; hardware; point selection ... See more keywords