Latency and accuracy optimization for binary neural network inference with locality‐aware operation skipping

• Published in: Electronics letters 2024-01, Vol.60 (2), p.n/a
• Main Authors: Lee, S.‐J., Kim, T.‐H.
• Format: Article
• Language: English
• Subjects: acceleration; artificial intelligence; inference mechanisms; neural net architecture
• ISSN: 0013-5194; 1350-911X
• DOI: 10.1049/ell2.13090

This letter presents a novel technique to optimize latency and accuracy for the inference based on binary neural network (BNN). The effects of the spatial locality in feature maps on latency and accuracy are analyzed in the BNN inference with the previous operation‐skipping method. A regularization‐based technique is proposed to adjust the locality with the aim of further reducing latency and improving accuracy of the previous method. In the CIFAR10 classification task, 11.62% latency reduction or 0.77% accuracy increase can be achieved when optimizing for each individually. When optimizing for both simultaneously, 5.58% latency reduction and 0.59% accuracy increase can be achieved. A regularization‐based technique is proposed to adjust the locality with the aim of reducing latency and improving accuracy. In the CIFAR10 classification task, the latency reduction rate is increased up to 40.20% by the proposed technique, which is 11.62% higher compared to that without employing the technique. The accuracy improvement is 0.77%.