Salvato in:
| Autori principali: | , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2019
|
| Soggetti: | |
| Accesso online: | https://arxiv.org/abs/1910.09876 |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866915972788518912 |
|---|---|
| author | Sanyal, Arnab Beerel, Peter A. Chugg, Keith M. |
| author_facet | Sanyal, Arnab Beerel, Peter A. Chugg, Keith M. |
| contents | The high computational complexity associated with training deep neural networks limits online and real-time training on edge devices. This paper proposed an end-to-end training and inference scheme that eliminates multiplications by approximate operations in the log-domain which has the potential to significantly reduce implementation complexity. We implement the entire training procedure in the log-domain, with fixed-point data representations. This training procedure is inspired by hardware-friendly approximations of log-domain addition which are based on look-up tables and bit-shifts. We show that our 16-bit log-based training can achieve classification accuracy within approximately 1% of the equivalent floating-point baselines for a number of commonly used datasets. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_1910_09876 |
| institution | arXiv |
| publishDate | 2019 |
| record_format | arxiv |
| spellingShingle | Neural Network Training with Approximate Logarithmic Computations Sanyal, Arnab Beerel, Peter A. Chugg, Keith M. Machine Learning The high computational complexity associated with training deep neural networks limits online and real-time training on edge devices. This paper proposed an end-to-end training and inference scheme that eliminates multiplications by approximate operations in the log-domain which has the potential to significantly reduce implementation complexity. We implement the entire training procedure in the log-domain, with fixed-point data representations. This training procedure is inspired by hardware-friendly approximations of log-domain addition which are based on look-up tables and bit-shifts. We show that our 16-bit log-based training can achieve classification accuracy within approximately 1% of the equivalent floating-point baselines for a number of commonly used datasets. |
| title | Neural Network Training with Approximate Logarithmic Computations |
| topic | Machine Learning |
| url | https://arxiv.org/abs/1910.09876 |