SPRING: A Sparsity-Aware Reduced-Precision Monolithic 3D CNN Accelerator Architecture for Training and Inference

Research output: Contribution to journalArticlepeer-review

4 Scopus citations


CNNs outperform traditional machine learning algorithms across a wide range of applications. However, their ever-growing computational complexity makes it necessary to design efficient hardware accelerators. The potential performance improvement from sparsity has not been adequately addressed. The computation and memory footprint of CNNs can be significantly reduced if sparsity is exploited in network evaluations. It has been shown that activations and weights also have high sparsity levels during the network training phase. Hence, sparsity-aware computation should also be considered in the training phase. To further improve performance and energy efficiency, some accelerators evaluate CNNs with limited precision. However, this is limited to the inference phase since reduced precision sacrifices network accuracy if used in training. In addition, CNN evaluation is usually memory-intensive, especially during training. In this article, we propose SPRING, a SParsity-aware Reduced-precision Monolithic 3D CNN accelerator for trainING and inference. It uses a binary mask scheme to encode sparsities. It uses the stochastic rounding algorithm to train CNNs with reduced precision without accuracy loss. To alleviate the memory bottleneck in CNN evaluation, especially during training, SPRING uses an efficient monolithic 3D nonvolatile memory interface to increase memory bandwidth.

Original languageEnglish (US)
JournalIEEE Transactions on Emerging Topics in Computing
StateAccepted/In press - 2020

All Science Journal Classification (ASJC) codes

  • Computer Science (miscellaneous)
  • Information Systems
  • Human-Computer Interaction
  • Computer Science Applications


  • Convolutional neural network
  • IC design
  • deep learning
  • hardware accelerator
  • inference
  • reduced precision
  • sparsity
  • stochastic rounding
  • training


Dive into the research topics of 'SPRING: A Sparsity-Aware Reduced-Precision Monolithic 3D CNN Accelerator Architecture for Training and Inference'. Together they form a unique fingerprint.

Cite this