Gated Parametric Neuron for Spike-based Audio Recognition
Haoran Wang, Herui Zhang, Siyang Li, Dongrui Wu · 2024
Read Gated Parametric Neuron for Spike-based Audio Recognition by Haoran Wang, Herui Zhang, Siyang Li, Dongrui Wu (2024) 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/arxiv-2412-01087.txt.
Description
Haoran Wang and co-authors present a 2024 neuron design for spiking neural networks that hear sound. Such networks mimic the brain by firing brief spikes rather than steady signals. The standard neuron for them fades its learning signal over long spans, which hurts training. Its settings are also fixed by hand, unlike real neurons that vary widely. The new design adds a gate that lets learning signals flow better. It also learns its own settings for each place and time.
The same gate removes the need to pick starting settings by hand. The team paired the neuron with a hybrid network that mixes two network styles. Tests on two spike-based sound data sets beat leading rival networks. The design did ease the fading signal problem as hoped. It also produced the varied settings seen across its neurons. The work targets tasks where sound unfolds over long spans.
Catalog record
| CREATOR | Haoran Wang, Herui Zhang, Siyang Li, Dongrui Wu |
|---|---|
| YEAR | 2024 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 35 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2412-01087