HEIF: Highly Efficient Stochastic Computing based Inference Framework for Deep Neural Networks

Zhe Li, Ji Li, Ao Ren, Ruizhe Cai, Caiwen Ding, Xuehai Qian, Jeffrey Draper, Bo Yuan, Jian Tang, Qinru Qiu, Yanzhi Wang

Research output: Contribution to journalArticle

4 Scopus citations

Abstract

Deep Convolutional Neural Networks (DCNNs) are one of the most promising types of deep learning technique and have been recognized as the dominant approach for almost all recognition and detection tasks. The computation of DCNNs is highly computational and memory intensive for the large feature maps and neuron connections, and the performance highly depends on the capability of hardware resources. With the recent trend of wearable devices and Internet of Things (IoTs), it becomes attractive to integrate the DCNNs onto embedded and portable devices, which require low power & energy consumptions and small hardware footprints. Recent work SC-DCNN asplos demonstrates that Stochastic Computing (SC), as a low-cost substitute to binary-based computing, can radically simplify the hardware implementation of arithmetic units and has the potential to satisfy the stringent power requirements in embedded devices. In SC, many arithmetic operations that are resource-consuming in binary designs can be implemented with very simple hardware logic, alleviating the extensive computational complexity. It offers a colossal design space for integration and optimization due to its reduced area and soft error resiliency. In this paper, we present HEIF, a highly efficient SC-based inference framework of the large-scale DCNNs, with broad applications including (but not limited to) LeNet-5 and AlexNet, that achieves high energy efficiency and low area/hardware cost. Compared to SC-DCNN asplos, HEIF features with 1) the first (to the best of our knowledge) SC-based Rectified Linear Unit (ReLU) activation function to catch up with the recent advances in software models and mitigate degradation in application-level accuracy; 2) the redesigned Approximate Parallel Counter (APC) and optimized stochastic multiplication using transmission gates and inverse mirror adders; and 3) the new optimization of weight storage using clustering. Most importantly, to achieve maximum energy efficiency while maintaining acceptable accuracy, HEIF considers holistic optimizations on cascade connection of function blocks in DCNN, pipelining technique, and bit-stream length reduction. Experimental results show that in large-scale applications HEIF outperforms previous SC-DCNN by the throughput of 4.1×, by area efficiency of up to 6.5× and achieves up to 5.6× energy improvement.

Keywords

  • ASIC
  • Convolutional Neural Network
  • Convolutional neural networks
  • Deep learning
  • Energy-efficient
  • Feature extraction
  • Hardware
  • Machine learning
  • Neurons
  • Optimization
  • Optimization.
  • Stochastic Computing

ASJC Scopus subject areas

  • Software
  • Computer Graphics and Computer-Aided Design
  • Electrical and Electronic Engineering

Fingerprint Dive into the research topics of 'HEIF: Highly Efficient Stochastic Computing based Inference Framework for Deep Neural Networks'. Together they form a unique fingerprint.

  • Cite this