Total for the last 12 months
number of access : ?
number of downloads : ?
ID 118049
Author
Nsinga, Robert Tokushima University
Keywords
embedded system
IEEE 754-2008 floating-point
digital signal processor
Q format notation
Content Type
Journal Article
Description
Using less electric power or speeding up processing is catching the interests of researchers in deep learning. Quantization has offered distillation mechanisms that substitute floating numbers for integers, but little has been suggested about the floating numbers themselves. The use of Q-format notation reduces computational overheads that frees resources for the introduction of more operations. Our experiments, conditioned on varying regimes, introduce automatic differentiation on algorithms like the fast Fourier transforms and Winograd minimal filtering to reduce computational complexity (expressed in total number of MACs) and suggest a path towards the assistive intelligence concept. Empirical results show that, under specific heuristics, the Q-format number notation can overcome the shortfalls of floating numbers, especially for embedded systems. Further benchmarks like the FPBench standard give more details by comparing our proposals with common deep learning operations.
Journal Title
Journal of Signal Processing
ISSN
18801013
Publisher
Research Institute of Signal Processing
Volume
26
Issue
5
Start Page
131
End Page
140
Published Date
2022-09-01
Remark
利用は著作権の範囲内に限られる。
EDB ID
DOI (Published Version)
URL ( Publisher's Version )
FullText File
language
eng
TextVersion
Publisher
departments
Science and Technology