An In-Memory-Computing Binary Neural Network Architecture With In-Memory Batch Normalization

This paper describes an in-memory computing architecture that combines full-precision computation for the first and last layers of a neural network while employing binary weights and input activations for the intermediate layers. This unique approach presents an efficient and effective solution for...

Full description

Saved in:
Bibliographic Details
Main Authors: Prathamesh Prashant Rege, Ming Yin, Sanjay Parihar, Joseph Versaggi, Shashank Nemawarkar
Format: Article
Language:English
Published: IEEE 2024-01-01
Series:IEEE Access
Subjects:
Online Access:https://ieeexplore.ieee.org/document/10639963/
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850254742931898368
author Prathamesh Prashant Rege
Ming Yin
Sanjay Parihar
Joseph Versaggi
Shashank Nemawarkar
author_facet Prathamesh Prashant Rege
Ming Yin
Sanjay Parihar
Joseph Versaggi
Shashank Nemawarkar
author_sort Prathamesh Prashant Rege
collection DOAJ
description This paper describes an in-memory computing architecture that combines full-precision computation for the first and last layers of a neural network while employing binary weights and input activations for the intermediate layers. This unique approach presents an efficient and effective solution for optimizing neural-network computations, reducing complexity, and enhancing energy efficiency. Notably, multiple architecture-level optimization methods are developed to ensure the binary operations thereby eliminating the need for intricate “digital logic” components external to the memory units. One of the key contributions of this study is in-memory batch normalization, which is implemented to provide good accuracy for CIFAR10 classification applications. Despite the inherent challenges posed by the process variations, the proposed design demonstrated an accuracy of 78%. Furthermore, the SRAM layer in the architecture showed an energy efficiency of 1086 TOPS/W and throughput of 23 TOPS, all packed efficiently within an area of 60 TOPS/mm2. This novel in-memory computing architecture offers a promising solution for next-generation efficient and high-performance deep learning applications.
format Article
id doaj-art-771dfe3f27e6483293c536acc86afa1d
institution OA Journals
issn 2169-3536
language English
publishDate 2024-01-01
publisher IEEE
record_format Article
series IEEE Access
spelling doaj-art-771dfe3f27e6483293c536acc86afa1d2025-08-20T01:57:04ZengIEEEIEEE Access2169-35362024-01-011219088919089610.1109/ACCESS.2024.344448110639963An In-Memory-Computing Binary Neural Network Architecture With In-Memory Batch NormalizationPrathamesh Prashant Rege0https://orcid.org/0009-0007-8244-2425Ming Yin1https://orcid.org/0009-0003-3916-0423Sanjay Parihar2https://orcid.org/0009-0001-8003-6956Joseph Versaggi3https://orcid.org/0009-0008-3117-7729Shashank Nemawarkar4Northeastern University, Boston, MA, USAGlobalFoundries, Malta, NY, USAGlobalFoundries, Austin, TX, USAGlobalFoundries, Malta, NY, USAGlobalFoundries, Austin, TX, USAThis paper describes an in-memory computing architecture that combines full-precision computation for the first and last layers of a neural network while employing binary weights and input activations for the intermediate layers. This unique approach presents an efficient and effective solution for optimizing neural-network computations, reducing complexity, and enhancing energy efficiency. Notably, multiple architecture-level optimization methods are developed to ensure the binary operations thereby eliminating the need for intricate “digital logic” components external to the memory units. One of the key contributions of this study is in-memory batch normalization, which is implemented to provide good accuracy for CIFAR10 classification applications. Despite the inherent challenges posed by the process variations, the proposed design demonstrated an accuracy of 78%. Furthermore, the SRAM layer in the architecture showed an energy efficiency of 1086 TOPS/W and throughput of 23 TOPS, all packed efficiently within an area of 60 TOPS/mm2. This novel in-memory computing architecture offers a promising solution for next-generation efficient and high-performance deep learning applications.https://ieeexplore.ieee.org/document/10639963/Batch normalizationbinary neural networkedge devicein-memory computingprocess variationSRAM
spellingShingle Prathamesh Prashant Rege
Ming Yin
Sanjay Parihar
Joseph Versaggi
Shashank Nemawarkar
An In-Memory-Computing Binary Neural Network Architecture With In-Memory Batch Normalization
IEEE Access
Batch normalization
binary neural network
edge device
in-memory computing
process variation
SRAM
title An In-Memory-Computing Binary Neural Network Architecture With In-Memory Batch Normalization
title_full An In-Memory-Computing Binary Neural Network Architecture With In-Memory Batch Normalization
title_fullStr An In-Memory-Computing Binary Neural Network Architecture With In-Memory Batch Normalization
title_full_unstemmed An In-Memory-Computing Binary Neural Network Architecture With In-Memory Batch Normalization
title_short An In-Memory-Computing Binary Neural Network Architecture With In-Memory Batch Normalization
title_sort in memory computing binary neural network architecture with in memory batch normalization
topic Batch normalization
binary neural network
edge device
in-memory computing
process variation
SRAM
url https://ieeexplore.ieee.org/document/10639963/
work_keys_str_mv AT prathameshprashantrege aninmemorycomputingbinaryneuralnetworkarchitecturewithinmemorybatchnormalization
AT mingyin aninmemorycomputingbinaryneuralnetworkarchitecturewithinmemorybatchnormalization
AT sanjayparihar aninmemorycomputingbinaryneuralnetworkarchitecturewithinmemorybatchnormalization
AT josephversaggi aninmemorycomputingbinaryneuralnetworkarchitecturewithinmemorybatchnormalization
AT shashanknemawarkar aninmemorycomputingbinaryneuralnetworkarchitecturewithinmemorybatchnormalization
AT prathameshprashantrege inmemorycomputingbinaryneuralnetworkarchitecturewithinmemorybatchnormalization
AT mingyin inmemorycomputingbinaryneuralnetworkarchitecturewithinmemorybatchnormalization
AT sanjayparihar inmemorycomputingbinaryneuralnetworkarchitecturewithinmemorybatchnormalization
AT josephversaggi inmemorycomputingbinaryneuralnetworkarchitecturewithinmemorybatchnormalization
AT shashanknemawarkar inmemorycomputingbinaryneuralnetworkarchitecturewithinmemorybatchnormalization