期刊
NATURE MACHINE INTELLIGENCE
卷 3, 期 3, 页码 -出版社
NATURE PORTFOLIO
DOI: 10.1038/s42256-021-00311-4
关键词
-
资金
- Human Brain Project of the European Union [785907]
Optimizing spiking neuron models for information transmission enhances the efficiency and accuracy of deep learning applications through reducing the number of spikes emitted per neuron. This new method improves latency and throughput of resulting spiking networks, offering a low-energy solution for edge and mobile devices in image classification tasks.
Spike-based neuromorphic hardware promises to reduce the energy consumption of image classification and other deep-learning applications, particularly on mobile phones and other edge devices. However, direct training of deep spiking neural networks is difficult, and previous methods for converting trained artificial neural networks to spiking neurons were inefficient because the neurons had to emit too many spikes. We show that a substantially more efficient conversion arises when one optimizes the spiking neuron model for that purpose, so that it not only matters for information transmission how many spikes a neuron emits, but also when it emits those spikes. This advances the accuracy that can be achieved for image classification with spiking neurons, and the resulting networks need on average just two spikes per neuron for classifying an image. In addition, our new conversion method improves latency and throughput of the resulting spiking networks. Spiking neural networks could offer a low-energy consuming solution to deep learning applications on the edge and in mobile devices. Using temporal coding, where the timing of spikes carries extra information, a new method efficiently converts conventional artificial neural networks to spiking networks.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据