An In-Memory-Computing Charge-Domain Ternary CNN Classifier
العنوان: | An In-Memory-Computing Charge-Domain Ternary CNN Classifier |
---|---|
المؤلفون: | Mingtao Zhan, Yongpan Liu, Xiyuan Tang, David Z. Pan, Keren Zhu, Jaydeep P. Kulkarni, Nan Sun, Meizhi Wang, Xiangxing Yang, Nanshu Lu |
المصدر: | CICC |
بيانات النشر: | Institute of Electrical and Electronics Engineers (IEEE), 2023. |
سنة النشر: | 2023 |
مصطلحات موضوعية: | Reduction (complexity), Artificial neural network, Edge device, Computer science, In-Memory Processing, Enhanced Data Rates for GSM Evolution, Electrical and Electronic Engineering, Convolutional neural network, Algorithm, MNIST database, Efficient energy use |
الوصف: | AI edge devices require local intelligence for the concerns of latency and privacy. Given the accuracy and energy constraints, low-power convolutional neural networks (CNNs) are gaining popularity. To alleviate the high memory access energy and computational cost of large CNN models, prior works have proposed promising approaches including in-memory-computing (IMC) [1], mixed-signal multiply-and-accumulate (MAC) calculation [2], and reduced resolution network –[4]. With weights and activations restricted to ±1, binary neural network (BNN) combining with IMC greatly improves the storage and computation efficiency, making it well-suited for edge-based applications, and has demonstrated state-of-the-art energy efficiency in image classification problems [5]. However, compared to full resolution network, BNN requires larger model thus more operations (OPs) per inference for a certain accuracy. To address such challenge, we propose a mixed-signal ternary CNN based processor featuring higher energy efficiency than BNN. It confers several key improvements: 1) the proposed ternary network provides 1.5-b resolution (0/+1/-1), leading to 3.9x OPs/inference reduction than BNN for the same MNIST accuracy; 2) a 1.5b MAC is implemented by V CM -based capacitor switching scheme, which inherently benefits from the reduced signal swing on the capacitive DAC (CDAC); 3) the V CM -based MAC introduces sparsity during training, resulting in lower switching rate. With a complete neural network on chip, the proposed design realizes 97.1% MNIST accuracy with only 0.18uJ per classification, presenting the highest power efficiency for comparable MNIST accuracy. |
تدمد: | 1558-173X 0018-9200 |
الوصول الحر: | https://explore.openaire.eu/search/publication?articleId=doi_dedup___::787c64a648b8098237df501f9c1ef935Test https://doi.org/10.1109/jssc.2023.3238725Test |
حقوق: | CLOSED |
رقم الانضمام: | edsair.doi.dedup.....787c64a648b8098237df501f9c1ef935 |
قاعدة البيانات: | OpenAIRE |
تدمد: | 1558173X 00189200 |
---|