Accelerating Digit Classification on FPGA with Pruned Binarized Neural Networks
Optimizations in Applied Machine Learning · 2024
DOI: 10.71070/oaml.v4i1.111 ↗Kurz gefasst
Fachartikel in der Zeitschrift Optimizations in Applied Machine Learning, 2024. Verfasst von Syamantak Payra und 3 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY-NC 4.0).
Abstract (Originalsprache)
As neural networks are increasingly deployed on mobile and distributed computing platforms, there is a need to lower latency and increase computational speed while decreasing power and memory usage. Rather than using FPGAs as accelerators in tandem with CPUs or GPUs, we directly encode individual neural network layers as combinational logic within FPGA hardware. Utilizing binarized neural networks minimizes the arithmetic computation required, shrinking latency to only the signal propagation delay. We evaluate size-optimization strategies and demonstrate network compression via weight quantization and weight-model unification, achieving 96% of the accuracy of baseline MNIST digit classification models while using only 3% of the memory. We further achieve 86% decrease in model footprint, 8mW dynamic power consumption, and
Zitieren
APA
Payra, S., Loke, G., Fink, Y., & Steinmeyer, J. D. (2024). Accelerating Digit Classification on FPGA with Pruned Binarized Neural Networks. Optimizations in Applied Machine Learning. https://doi.org/10.71070/oaml.v4i1.111
Vancouver
Payra S, Loke G, Fink Y, Steinmeyer JD. Accelerating Digit Classification on FPGA with Pruned Binarized Neural Networks. Optimizations in Applied Machine Learning. 2024. doi:10.71070/oaml.v4i1.111
BibTeX
@article{Payra2024,
author = {Payra, Syamantak and Loke, Gabriel and Fink, Yoel and Steinmeyer, Joseph D.},
title = {Accelerating Digit Classification on FPGA with Pruned Binarized Neural Networks},
journal = {Optimizations in Applied Machine Learning},
year = {2024},
doi = {10.71070/oaml.v4i1.111}
}
EUVIMED