Neuron pruning in temporal domain for energy efficient SNN processor design
Dongwoo Lew, Hoyoung Tang, Jongsun Park · 2023
Read Neuron pruning in temporal domain for energy efficient SNN processor design by Dongwoo Lew, Hoyoung Tang, Jongsun Park (2023) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/pmc-10719842.txt.
Description
Dongwoo Lew, Hoyoung Tang, and Jongsun Park present a pruning method for spiking neural network chips in a 2023 paper. Spiking networks have grown more accurate by borrowing designs from standard networks, but they demand heavy computation. That load blocks energy efficient chip design. Their method spots less important neurons during operation and skips their future work.
Pruning choices rest on voltage thresholds set layer by layer, found by two search methods that trade accuracy against savings. A chip built in a common process cut energy use by fifty seven percent and ran more than twice as fast. Accuracy fell by less than one percent on a standard image set. Readers see time domain pruning as a path to greener brain like computing.
Catalog record
| CREATOR | Dongwoo Lew, Hoyoung Tang, Jongsun Park |
|---|---|
| YEAR | 2023 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 42 KB |
| VERSION | 1.0 |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |