Integration of Ag-CBRAM crossbars and Mott ReLU neurons for efficient implementation of deep neural networks in hardware
Abstract
Abstract In-memory computing with emerging non-volatile memory devices (eNVMs) has shown promising results in accelerating matrix-vector multiplications (MVMs). However, activation function calculations are still being implemented with general processors or large and complex neuron peripheral circuits. Here, we present the integration of Ag-based conductive bridge random access memory (Ag-CBRAM) crossbar arrays with Mott ReLU activation neurons for scalable, energy and area-efficient hardware implementation of deep neural networks (DNNs). We develop Ag-CBRAM devices that can achieve a high ON/OFF ratio and multi-level programmability. Compact and energy-efficient Mott ReLU neuron devices implementing rectified linear unit (ReLU) activation function are directly connected to the columns of Ag-CBRAM crossbars to compute the output from the weighted sum current. We implement convolution filters and activations for VGG-16 using our integrated hardware and demonstrate the successful generation of feature maps for CIFAR-10 images in hardware. Our approach paves a new way toward building a highly compact and energy-efficient eNVMs-based in-memory computing system.
- Authors:
- Publication Date:
- Sponsoring Org.:
- USDOE
- OSTI Identifier:
- 1997290
- Alternate Identifier(s):
- OSTI ID: 1994436
- Grant/Contract Number:
- #DE-SC0019273
- Resource Type:
- Published Article
- Journal Name:
- Neuromorphic Computing and Engineering
- Additional Journal Information:
- Journal Name: Neuromorphic Computing and Engineering Journal Volume: 3 Journal Issue: 3; Journal ID: ISSN 2634-4386
- Publisher:
- IOP Publishing
- Country of Publication:
- United Kingdom
- Language:
- English
Citation Formats
Shi, Yuhan, Oh, Sangheon, Park, Jaeseoung, Valle, Javier del, Salev, Pavel, Schuller, Ivan K., and Kuzum, Duygu. Integration of Ag-CBRAM crossbars and Mott ReLU neurons for efficient implementation of deep neural networks in hardware. United Kingdom: N. p., 2023.
Web. doi:10.1088/2634-4386/aceea9.
Shi, Yuhan, Oh, Sangheon, Park, Jaeseoung, Valle, Javier del, Salev, Pavel, Schuller, Ivan K., & Kuzum, Duygu. Integration of Ag-CBRAM crossbars and Mott ReLU neurons for efficient implementation of deep neural networks in hardware. United Kingdom. https://doi.org/10.1088/2634-4386/aceea9
Shi, Yuhan, Oh, Sangheon, Park, Jaeseoung, Valle, Javier del, Salev, Pavel, Schuller, Ivan K., and Kuzum, Duygu. Tue .
"Integration of Ag-CBRAM crossbars and Mott ReLU neurons for efficient implementation of deep neural networks in hardware". United Kingdom. https://doi.org/10.1088/2634-4386/aceea9.
@article{osti_1997290,
title = {Integration of Ag-CBRAM crossbars and Mott ReLU neurons for efficient implementation of deep neural networks in hardware},
author = {Shi, Yuhan and Oh, Sangheon and Park, Jaeseoung and Valle, Javier del and Salev, Pavel and Schuller, Ivan K. and Kuzum, Duygu},
abstractNote = {Abstract In-memory computing with emerging non-volatile memory devices (eNVMs) has shown promising results in accelerating matrix-vector multiplications (MVMs). However, activation function calculations are still being implemented with general processors or large and complex neuron peripheral circuits. Here, we present the integration of Ag-based conductive bridge random access memory (Ag-CBRAM) crossbar arrays with Mott ReLU activation neurons for scalable, energy and area-efficient hardware implementation of deep neural networks (DNNs). We develop Ag-CBRAM devices that can achieve a high ON/OFF ratio and multi-level programmability. Compact and energy-efficient Mott ReLU neuron devices implementing rectified linear unit (ReLU) activation function are directly connected to the columns of Ag-CBRAM crossbars to compute the output from the weighted sum current. We implement convolution filters and activations for VGG-16 using our integrated hardware and demonstrate the successful generation of feature maps for CIFAR-10 images in hardware. Our approach paves a new way toward building a highly compact and energy-efficient eNVMs-based in-memory computing system.},
doi = {10.1088/2634-4386/aceea9},
journal = {Neuromorphic Computing and Engineering},
number = 3,
volume = 3,
place = {United Kingdom},
year = {Tue Aug 29 00:00:00 EDT 2023},
month = {Tue Aug 29 00:00:00 EDT 2023}
}
https://doi.org/10.1088/2634-4386/aceea9
Works referenced in this record:
Wafer-scale growth of VO2 thin films using a combinatorial approach
journal, October 2015
- Zhang, Hai-Tian; Zhang, Lei; Mukherjee, Debangshu
- Nature Communications, Vol. 6, Issue 1
Metal-Insulator Transition in ALD VO 2 Ultrathin Films and Nanoparticles: Morphological Control
journal, December 2014
- Peter, Antony P.; Martens, Koen; Rampelberg, Geert
- Advanced Functional Materials, Vol. 25, Issue 5
Deep learning
journal, May 2015
- LeCun, Yann; Bengio, Yoshua; Hinton, Geoffrey
- Nature, Vol. 521, Issue 7553
A comparison of deep networks with ReLU activation function and linear spline-type methods
journal, February 2019
- Eckle, Konstantin; Schmidt-Hieber, Johannes
- Neural Networks, Vol. 110
Learning in Memristive Neural Network Architectures Using Analog Backpropagation Circuits
journal, February 2019
- Krestinskaya, Olga; Salama, Khaled Nabil; James, Alex Pappachen
- IEEE Transactions on Circuits and Systems I: Regular Papers, Vol. 66, Issue 2
Analog-to-Digital Conversion With Reconfigurable Function Mapping for Neural Networks Activation Function Acceleration
journal, June 2019
- Giordano, Massimo; Cristiano, Giorgio; Ishibashi, Koji
- IEEE Journal on Emerging and Selected Topics in Circuits and Systems, Vol. 9, Issue 2
Heterogeneous integration of single-crystalline rutile nanomembranes with steep phase transition on silicon substrates
journal, August 2021
- Lee, Dong Kyu; Park, Yunkyu; Sim, Hyeji
- Nature Communications, Vol. 12, Issue 1
In-memory computing with resistive switching devices
journal, June 2018
- Ielmini, Daniele; Wong, H. -S. Philip
- Nature Electronics, Vol. 1, Issue 6
Compact Modeling of RRAM Devices and Its Applications in 1T1R and 1S1R Array Design
journal, December 2015
- Chen, Pai-Yu; Yu, Shimeng
- IEEE Transactions on Electron Devices, Vol. 62, Issue 12
Monolithically Integrated RRAM- and CMOS-Based In-Memory Computing Optimizations for Efficient Deep Learning
journal, November 2019
- Yin, Shihui; Seo, Jae-sun; Kim, Yulhwa
- IEEE Micro, Vol. 39, Issue 6
Oxide-based filamentary RRAM for deep learning
journal, December 2020
- Zhang, Yizhou; Huang, Peng; Gao, Bin
- Journal of Physics D: Applied Physics, Vol. 54, Issue 8
Research progress on solutions to the sneak path issue in memristor crossbar arrays
journal, January 2020
- Shi, Lingyun; Zheng, Guohao; Tian, Bobo
- Nanoscale Advances, Vol. 2, Issue 5
Energy-efficient Mott activation neuron for full-hardware implementation of neural networks
journal, March 2021
- Oh, Sangheon; Shi, Yuhan; del Valle, Javier
- Nature Nanotechnology, Vol. 16, Issue 6
Reconfigurable anisotropy and functional transformations with -based metamaterial electric circuits
journal, April 2015
- Savo, Salvatore; Zhou, You; Castaldi, Giuseppe
- Physical Review B, Vol. 91, Issue 13
Conduction mechanism of a TaO x -based selector and its application in crossbar memory arrays
journal, January 2015
- Wang, Ming; Zhou, Jiantao; Yang, Yuchao
- Nanoscale, Vol. 7, Issue 11