Skip to main content
U.S. Department of Energy
Office of Scientific and Technical Information

Analog architectures for neural network acceleration based on non-volatile memory

Journal Article · · Applied Physics Reviews
DOI:https://doi.org/10.1063/1.5143815· OSTI ID:2564777

Analog hardware accelerators, which perform computation within a dense memory array, have the potential to overcome the major bottlenecks faced by digital hardware for data-heavy workloads such as deep learning. Exploiting the intrinsic computational advantages of memory arrays, however, has proven to be challenging principally due to the overhead imposed by the peripheral circuitry and due to the non-ideal properties of memory devices that play the role of the synapse. In this paper, we review the existing implementations of these accelerators for deep supervised learning, organizing our discussion around the different levels of the accelerator design hierarchy, with an emphasis on circuits and architecture. We also explore and consolidate the various approaches that have been proposed to address the critical challenges faced by analog accelerators, for both neural network inference and training, and highlight the key design tradeoffs underlying these techniques.

Research Organization:
Sandia National Lab. (SNL-NM), Albuquerque, NM (United States)
Sponsoring Organization:
USDOE National Nuclear Security Administration (NNSA); USDOE Laboratory Directed Research and Development (LDRD) Program
Grant/Contract Number:
AC04-94AL85000; NA0003525
OSTI ID:
2564777
Alternate ID(s):
OSTI ID: 1637994; OSTI ID: 1668360
Report Number(s):
SAND-2020-5154J; 686136; TRN: US2203792
Journal Information:
Applied Physics Reviews, Vol. 7, Issue 3; ISSN 1931-9401
Publisher:
American Institute of Physics (AIP)Copyright Statement
Country of Publication:
United States
Language:
English

References (130)

Why is nonvolatile ferroelectric memory field-effect transistor still elusive? journal July 2002
Mixed-Precision Deep Learning Based on Computational Memory journal May 2020
3D Ta/TaO x /TiO 2 /Ti synaptic array and linearity tuning of weight update for hardware neural network applications journal August 2016
Repeatable, accurate, and high speed multi-level programming of memristor 1T1R arrays for power efficient analog computing applications journal August 2016
Parallel Architecture With Resistive Crosspoint Array for Dictionary Learning Acceleration journal June 2015
Super non-linear RRAM with ultra-low power for 3D vertical nano-crossbar arrays journal January 2016
Redesigning commercial floating-gate memory for analog computing applications conference May 2015
Memristor-Based Analog Computation and Neural Network Classification with a Dot Product Engine journal January 2018
Large-scale neural networks implemented with non-volatile memory as the synaptic weight element: Comparative performance analysis (accuracy, speed, and power) conference December 2015
Efficient and self-adaptive in-situ learning in multilayer memristor neural networks journal June 2018
Access devices for 3D crosspoint memory
  • Burr, Geoffrey W.; Shenoy, Rohit S.; Virwani, Kumar
  • Journal of Vacuum Science & Technology B, Nanotechnology and Microelectronics: Materials, Processing, Measurement, and Phenomena, Vol. 32, Issue 4 https://doi.org/10.1116/1.4889999
journal July 2014
Memristor-Based Multilayer Neural Networks With Online Gradient Descent Training journal October 2015
Newton: Gravitating Towards the Physical Limits of Crossbar Acceleration journal September 2018
Finding a roadmap to achieve large neuromorphic hardware systems journal January 2013
Pattern classification by memristive crossbar circuits using ex situ and in situ training journal June 2013
Vector-Matrix Multiply and Winner-Take-All as an Analog Classifier journal February 2014
Computation-oriented fault-tolerance schemes for RRAM computing systems conference January 2017
UNPU: A 50.6TOPS/W unified deep neural network accelerator with 1b-to-16b fully-variable weight bit-precision
  • Lee, Jinmook; Kim, Changhyeon; Kang, Sanghoon
  • 2018 IEEE International Solid-State Circuits Conference (ISSCC), 2018 IEEE International Solid - State Circuits Conference - (ISSCC) https://doi.org/10.1109/ISSCC.2018.8310262
conference February 2018
Improved Synaptic Behavior Under Identical Pulses Using AlO x /HfO 2 Bilayer RRAM Array for Neuromorphic Systems journal August 2016
A non-volatile organic electrochemical device as a low-voltage artificial synapse for neuromorphic computing journal February 2017
3-D Stacked Synapse Array Based on Charge-Trap Flash Memory for Implementation of Deep Neural Networks journal January 2019
Overview of candidate device technologies for storage-class memory journal July 2008
High-Performance Mixed-Signal Neurocomputing With Nanoscale Floating-Gate Memory Cell Arrays journal October 2018
24.2 A 2.5GHz 7.7TOPS/W switched-capacitor matrix multiplier with co-designed local memory in 40nm conference January 2016
Binary convolutional neural network on RRAM conference January 2017
Experimental Demonstration of Feature Extraction and Dimensionality Reduction Using Memristor Networks journal April 2017
Ferroelectric FETs-Based Nonvolatile Logic-in-Memory Circuits journal January 2019
Random synaptic feedback weights support error backpropagation for deep learning journal November 2016
Threshold-alterable Si-gate MOS devices journal May 1977
Long Short-Term Memory journal November 1997
Reducing the Impact of Phase-Change Memory Conductance Drift on the Inference of large-scale Hardware Neural Networks conference December 2019
Efficient Processing of Deep Neural Networks: A Tutorial and Survey journal December 2017
Using Floating-Gate Memory to Train Ideal Accuracy Neural Networks journal June 2019
The Next Generation of Deep Learning Hardware: Analog Computing journal January 2019
An analog neural network processor with programmable topology journal January 1991
Serving DNNs in Real Time at Datacenter Scale with Project Brainwave journal March 2018
Analog Architecture Complexity Theory Empowering Ultra-Low Power Configurable Analog and Mixed Mode SoC Systems journal January 2019
Analog-to-Digital Conversion With Reconfigurable Function Mapping for Neural Networks Activation Function Acceleration journal June 2019
A Reconfigurable Fabric for Accelerating Large-Scale Datacenter Services journal May 2015
Neuromorphic computing using non-volatile memory journal October 2016
Visual Pattern Extraction Using Energy-Efficient “2-PCM Synapse” Neuromorphic Architecture journal August 2012
A single-transistor silicon synapse journal January 1996
A Ferroelectric FET-Based Processing-in-Memory Architecture for DNN Acceleration journal December 2019
Metal–Oxide RRAM journal June 2012
A Programmable and Configurable Mixed-Mode FPAA SoC journal January 2016
A Survey of ReRAM-Based Architectures for Processing-In-Memory and Neural Networks journal April 2018
Binary neural network with 16 Mb RRAM macro chip for classification and online training conference December 2016
Gradient-based learning applied to document recognition journal January 1998
Charge-mode parallel architecture for vector-matrix multiplication
  • Genov, R.; Cauwenberghs, G.
  • IEEE Transactions on Circuits and Systems II: Analog and Digital Signal Processing, Vol. 48, Issue 10 https://doi.org/10.1109/82.974781
journal January 2001
PUMA: A Programmable Ultra-efficient Memristor-based Accelerator for Machine Learning Inference
  • Ankit, Aayush; Hajj, Izzat El; Chalamalasetti, Sai Rahul
  • ASPLOS '19: Architectural Support for Programming Languages and Operating Systems, Proceedings of the Twenty-Fourth International Conference on Architectural Support for Programming Languages and Operating Systems https://doi.org/10.1145/3297858.3304049
conference April 2019
An Energy-Efficient Digital ReRAM-Crossbar-Based CNN With Bitwise Parallelism journal December 2017
Energy-Efficient Time-Domain Vector-by-Matrix Multiplier for Neurocomputing and Beyond journal September 2019
Toward on-chip acceleration of the backpropagation algorithm using nonvolatile memory journal July 2017
Resistive memory device requirements for a neural algorithm accelerator conference July 2016
Analysis of Power Consumption and Linearity in Capacitive Digital-to-Analog Converters Used in Successive Approximation ADCs journal August 2011
Parasitic Effect Analysis in Memristor-Array-Based Neuromorphic Systems journal January 2018
BRein Memory: A Single-Chip Binary/Ternary Reconfigurable in-Memory Deep Neural Network Accelerator Achieving 1.4 TOPS at 0.6 W journal April 2018
Fully parallel write/read in resistive synaptic array for accelerating on-chip learning journal October 2015
Programmable analog vector-matrix multipliers journal January 1990
Kerneltron: support vector "machine" in silicon journal September 2003
Neuro-Inspired Computing With Emerging Nonvolatile Memorys journal February 2018
Anatomy of Ag/Hafnia-Based Selectors with 10 10 Nonlinearity journal January 2017
Enabling Scientific Computing on Memristive Accelerators conference June 2018
A Highly Dense, Low Power, Programmable Analog Vector-Matrix Multiplier: The FPAA Implementation journal September 2011
A fully integrated reprogrammable memristor–CMOS system for efficient multiply–accumulate operations journal July 2019
Harmonica: A Framework of Heterogeneous Computing Systems With Memristor-Based Neuromorphic Computing Accelerators journal May 2016
Hitting the memory wall: implications of the obvious journal March 1995
Brain-inspired computing with memristors: Challenges in devices, circuits, and systems journal March 2020
Analog Implementation of a Novel Resistive-Type Sigmoidal Neuron journal April 2012
A dynamically configurable coprocessor for convolutional neural networks conference January 2010
Distributed In-Memory Computing on Binary RRAM Crossbar journal May 2017
Optimal unsupervised learning in a single-layer linear feedforward neural network journal January 1989
A magnetic synapse: multilevel spin-torque memristor with perpendicular anisotropy journal August 2016
Implementation of multilayer perceptron network with highly uniform passive memristive crossbar circuits journal June 2018
A review of emerging non-volatile memory (NVM) technologies and applications journal November 2016
ImageNet Large Scale Visual Recognition Challenge journal April 2015
3-D Memristor Crossbars for Analog and Neuromorphic Computing Applications journal January 2017
Redox transistors for neuromorphic computing journal November 2019
New Memristor-Based Crossbar Array Architecture with 50-% Area Reduction and 48-% Power Saving for Matrix-Vector Multiplication of Analog Neuromorphic Computing journal June 2014
Scaling-up resistive synaptic arrays for neuro-inspired architecture: Challenges and prospect conference December 2015
Parallel programming of an ionic floating-gate memory array for scalable neuromorphic computing journal April 2019
Memristive devices for computing journal January 2013
Training LSTM Networks With Resistive Cross-Point Devices journal October 2018
Algorithm for Training Neural Networks on Resistive Device Arrays journal February 2020
RRAM-Based Analog Approximate Computing journal December 2015
Computational phase-change memory: beyond von Neumann computing journal August 2019
Acceleration of Deep Neural Network Training with Resistive Cross-Point Devices: Design Considerations journal July 2016
An approximate backpropagation learning rule for memristor based neural networks using synaptic plasticity journal May 2017
Li-Ion Synaptic Transistor for Low Power Analog Computing journal November 2016
Design Considerations for Efficient Deep Neural Networks on Processing-in-Memory Accelerators conference December 2019
Ferroelectric FET analog synapse for acceleration of deep neural network training conference December 2017
Training and operation of an integrated neuromorphic network based on metal-oxide memristors journal May 2015
High precision tuning of state for memristive devices by adaptable variation-tolerant algorithm journal January 2012
Ambit: in-memory accelerator for bulk bitwise operations using commodity DRAM technology conference January 2017
Chaotic dynamics in nanoscale NbO2 Mott memristors for analogue computing journal August 2017
A multiply-add engine with monolithically integrated 3D memristor crossbar/CMOS hybrid circuit journal February 2017
Physical Realization of a Supervised Learning System Built with Organic Memristive Synapses journal September 2016
A domain-specific architecture for deep neural networks journal August 2018
Bidirectional Non-Filamentary RRAM as an Analog Neuromorphic Synapse, Part I: Al/Mo/Pr 0.7 Ca 0.3 MnO 3 Material Improvements and Device Measurements journal January 2018
Recent progress in analog memory-based accelerators for deep learning journal June 2018
TiO x -Based RRAM Synapse With 64-Levels of Conductance and Symmetric Conductance Change by Adopting a Hybrid Pulse Scheme for Neuromorphic Computing journal December 2016
Experimental Demonstration and Tolerancing of a Large-Scale Neural Network (165 000 Synapses) Using Phase-Change Memory as the Synaptic Weight Element journal November 2015
Solving matrix equations in one step with cross-point resistive arrays journal February 2019
A fully analog memristor-based neural network with online gradient training conference May 2016
An Always-On 3.8 $\mu$ J/86% CIFAR-10 Mixed-Signal Binary CNN Processor With All Memory on Chip in 28-nm CMOS journal January 2019
Contrasting Advantages of Learning With Random Weights and Backpropagation in Non-Volatile Memory Neural Networks journal January 2019
DianNao: a small-footprint high-throughput accelerator for ubiquitous machine-learning
  • Chen, Tianshi; Du, Zidong; Sun, Ninghui
  • Proceedings of the 19th international conference on Architectural support for programming languages and operating systems - ASPLOS '14 https://doi.org/10.1145/2541940.2541967
conference January 2014
Linear electronic analog/digital conversion architectures, their origins, parameters, limitations, and applications journal July 1978
Neuromorphic computing with multi-memristive synapses journal June 2018
An Open-Source Tool Set Enabling Analog-Digital-Software Co-Design journal February 2016
Three-Dimensional nand Flash for Vector–Matrix Multiplication journal April 2019
Analog high resistance bilayer RRAM device for hardware acceleration of neuromorphic computation journal November 2018
Scaling for edge inference of deep neural networks journal April 2018
Face classification using electronic synapses journal May 2017
28.5 A 10b 1.5GS/s pipelined-SAR ADC with background second-stage common-mode regulation and offset calibration in 14nm CMOS FinFET
  • Kull, Lukas; Luu, Danny; Menolfi, Christian
  • 2017 IEEE International Solid- State Circuits Conference - (ISSCC), 2017 IEEE International Solid-State Circuits Conference (ISSCC) https://doi.org/10.1109/ISSCC.2017.7870467
conference February 2017
DRISA: a DRAM-based Reconfigurable In-Situ Accelerator conference January 2017
Multiscale Co-Design Analysis of Energy, Latency, Area, and Accuracy of a ReRAM Analog Neural Training Accelerator journal March 2018
Efficient Mixed-Signal Neurocomputing Via Successive Integration and Rescaling journal March 2020
Fully hardware-implemented memristor convolutional neural network journal January 2020
Monolithically Integrated RRAM- and CMOS-Based In-Memory Computing Optimizations for Efficient Deep Learning journal November 2019
Fast, energy-efficient, robust, and reproducible mixed-signal neuromorphic classifier based on embedded NOR flash memory technology conference December 2017
Equivalent-accuracy accelerated neural-network training using analogue memory journal June 2018
A survey on nonlinear analog-to-digital converters journal January 2014
All-Solid-State Synaptic Transistor with Ultralow Conductance for Neuromorphic Computing journal September 2018
Streaming Batch Eigenupdates for Hardware Neural Networks journal August 2019
YodaNN: An Ultra-Low Power Convolutional Neural Network Accelerator Based on Binary Weights conference July 2016
$\hbox{Al}_{2}\hbox{O}_{3}$-Based RRAM Using Atomic Layer Deposition (ALD) With 1-$\mu\hbox{A}$ RESET Current journal December 2010
Energy Scaling Advantages of Resistive Memory Crossbar Based Computation and Its Application to Sparse Coding journal January 2016
Learning the sparsity for ReRAM: mapping and pruning sparse neural network for ReRAM based accelerator
  • Lin, Jilan; Zhu, Zhenhua; Wang, Yu
  • ASPDAC '19: 24th Asia and South Pacific Design Automation Conference, Proceedings of the 24th Asia and South Pacific Design Automation Conference https://doi.org/10.1145/3287624.3287715
conference January 2019
Design of Ternary Neural Network With 3-D Vertical RRAM Array journal June 2017

Similar Records

Related Subjects