期刊
IEEE JOURNAL OF SOLID-STATE CIRCUITS
卷 52, 期 10, 页码 2679-2689出版社
IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
DOI: 10.1109/JSSC.2017.2712626
关键词
Analog computing; binarized neural network (BNN); convolutional neural network (CNN); deep learning; neuromorphic computing; time domain
Demand for highly energy-efficient coprocessor for the inference computation of deep neural networks is increasing. We propose the time-domain neural network (TDNN), which employs time-domain analog and digital mixed-signal processing (TDAMS) that uses delay time as the analog signal. TDNN not only exploits energy-efficient analog computing, but also enables fully spatially unrolled architecture by the hardwareefficient feature of TDAMS. The proposed fully spatially unrolled architecture reduces energy-hungry data moving for weight and activations, thus contributing to significant improvement of energy efficiency. We also propose useful training techniques that mitigate the non-ideal effect of analog circuits, which enables to simplify the circuits and leads to maximizing the energy efficiency. The proof-of-concept chip shows unprecedentedly high energy efficiency of 48.2 TSop/s/W.
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