Details

Autor(en) / Beteiligte

• Xia, Zihan
• Chen, Jienan
• He, Shaoxia
• Li, Shuai

Titel

Neural Synaptic Plasticity-Like Computing: An Ultra-Low Cost Approach for Artificial Neural Networks Implementation

Ist Teil von

• 2020 IEEE International Symposium on Circuits and Systems (ISCAS), 2020, p.1-5

Ort / Verlag

IEEE

Erscheinungsjahr

2020

Quelle

IEEE/IET Electronic Library (IEL)

Beschreibungen/Notizen

• Artificial neural networks (ANNs) have gained state-of-the-art results in classification and regression tasks. However, there is still great gap between ANNs and human brain in terms of computation efficiency. In this work, we proposed the neural synaptic plasticity-like computing (NSPC) to simulate the neural network activity for inference task with ultra-simple logic gates. The multiplication of weight in traditional ANNs is transformed by the wire connectivity in NSPC, which requires only bundle of wires without any logics. To this end, the NSPC imitates the structure of neural synaptic plasticity from a circuit wires connection perspective. The proposed NSPC exhibits comparable inference accuracy with low hardware cost. According to the implementation results, the NSPC requires only 28% logic gate resources of conventional ANNs scheme, 114% throughput improvement and 8.454 times better hardware efficiency on the average.