Fully Integrated Analog Machine Learning Classifier Using Custom Activation Function for Low Resolution Image Classification
Fully Integrated Analog Machine Learning Classifier Using Custom Activation Function for Low Resolution Image Classification
复制标题
DOI：
10.1109/tcsi.2020.3047331
复制
发表时间：
2021-03-01
影响因子：
5.1
通讯作者：
Sanyal, Arindam
中科院分区：
文献类型：
--
作者：
Chandrasekaran, Sanjeev Tannirkulam;Jayaraj, Akshay;Sanyal, Arindam
本文提出了一种完全集成的模拟神经网络分类器结构，用于低分辨率图像分类，消除了内存访问。我们使用单级共源极放大器设计自定义激活函数，并应用硬件-软件协同设计方法将自定义激活函数的知识纳入训练阶段，以实现高精度。完全在模拟域中执行所有计算消除了与存储器访问和数据移动相关联的能量成本。我们证明了我们的分类器识别MNIST数据集的下采样手写数字的多项式分类任务。该芯片采用65 nm CMOS工艺制造，下采样MNIST数据集的实测能耗为173 pJ/分类，比现有技术提高了3倍。即使在将原始MNIST图像从28 x 28像素下采样96%到5 x 5像素后，该芯片的平均分类精度仍达到81.3%。
This paper presents fully-integrated analog neural network classifier architecture for low resolution image classification that eliminates memory access. We design custom activation functions using single-stage common-source amplifiers, and apply a hardware-software co-design methodology to incorporate knowledge of the custom activation functions into the training phase to achieve high accuracy. Performing all computations entirely in the analog domain eliminates energy cost associated with memory access and data movement. We demonstrate our classifier on multinomial classification task of recognizing downsampled handwritten digits from MNIST dataset. Fabricated in 65nm CMOS process, the measured energy consumption for down-sampled MNIST dataset is 173pJ/classification, which is 3x better than state-of-the-art. The prototype IC achieves mean classification accuracy of 81.3% even after down-sampling the original MNIST images by 96% from 28 x 28 pixels to 5 x 5 pixels.