https://scholars.lib.ntu.edu.tw/handle/123456789/607214
標題: | Logic Synthesis of Binarized Neural Networks for Efficient Circuit Implementation | 作者: | Chi C JIE-HONG JIANG |
關鍵字: | binarized neural network;Biological neural networks;Complexity theory;Convolution;Integrated circuit interconnections;Integrated circuit modeling;Logic gates;logic synthesis;matrix covering.;model pruning;Neurons;Computer circuits;Deep learning;Learning systems;Logic circuits;Logic Synthesis;Memory architecture;Timing circuits;Circuit implementation;FPGA implementations;Hardware architecture;Hardware implementations;Hardware realization;Neural networks (NNS);Neural-processing;Performance efficiency;Neural networks | 公開日期: | 2021 | 來源出版物: | IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems | 摘要: | Neural networks (NNs) are key to deep learning systems. Their efficient hardware implementation is crucial to applications at the edge. Binarized NNs (BNNs), where the weights and output of a neuron are of binary values {-1,+1❵ (or encoded in {0,1❵), have been proposed. As no multiplier required, binarized neural networks are particularly attractive and suitable for hardware realization. Most prior NN synthesis methods target on hardware architectures with neural processing elements (NPEs), where the weights of a neuron are loaded and the output of the neuron is computed. The load-and-compute method, though area efficient, requires expensive memory access, which deteriorates energy and performance efficiency. In this work we aim at synthesizing BNN layers into dedicated logic circuits. We formulate the corresponding model pruning problem and matrix covering problem to reduce the area and routing cost of BNNs. For model pruning, we propose and compare three strategies at the BNN training stage. For matrix covering, we propose a scalable logic-sharing algorithm. By combining these two methods, experimental results justify the effectiveness of the method in terms of area and net savings on FPGA implementation. Our method provides an alternative implementation of BNNs, and can be applied in combination with NPE-based implementation for area, speed, and power tradeoffs. IEEE |
URI: | https://www.scopus.com/inward/record.uri?eid=2-s2.0-85105850641&doi=10.1109%2fTCAD.2021.3078606&partnerID=40&md5=173402d241fe42ce19b3dfa5e4579a4d https://scholars.lib.ntu.edu.tw/handle/123456789/607214 |
ISSN: | 02780070 | DOI: | 10.1109/TCAD.2021.3078606 |
顯示於: | 電機工程學系 |
在 IR 系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。