Convolutional Networks for Fast, Energy-Efficient Neuromorphic Computing

Abstract

Deep networks are now able to achieve human-level performance on a broad spectrum of recognition tasks. Independently, neuromorphic computing has now demonstrated unprecedented energy-efficiency through a new chip architecture based on spiking neurons, low precision synapses, and a scalable communication network. Here, we demonstrate that neuromorphic computing, despite its novel architectural primitives, can implement deep convolution networks that i) approach state-of-the-art classification accuracy across 8 standard datasets, encompassing vision and speech, ii) perform inference while preserving the hardware's underlying energy-efficiency and high throughput, running on the aforementioned datasets at between 1200 and 2600 frames per second and using between 25 and 275 mW (effectively > 6000 frames / sec / W) and iii) can be specified and trained using backpropagation with the same ease-of-use as contemporary deep learning. For the first time, the algorithmic power of deep learning can be merged with the efficiency of neuromorphic processors, bringing the promise of embedded, intelligent, brain-inspired computing one step closer.

7 pages, 6 figures

References in corpus (6)

• Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
• Striving for Simplicity: The All Convolutional Net
• BinaryConnect: Training Deep Neural Networks with binary weights during propagations
• Six networks on a universal neuromorphic computing substrate
• Adjustable Bounded Rectifiers: Towards Deep Binary Representations
• Gibbs Sampling with Low-Power Spiking Digital Neurons

Cited by in corpus (128)

• Deep Learning with Coherent Nanophotonic Circuits
• Event-based Vision: A Survey
• 11 TeraFLOPs per second photonic convolutional accelerator for deep learning optical neural networks
• Deep Learning in Spiking Neural Networks
• Spatio-Temporal Backpropagation for Training High-performance Spiking Neural Networks
• Ternary Weight Networks
• A scalable multi-core architecture with heterogeneous memory structures for Dynamic Neuromorphic Asynchronous Processors (DYNAPs)
• Enabling Spike-based Backpropagation for Training Deep Neural Network Architectures
• A Survey of Neuromorphic Computing and Neural Networks in Hardware
• Synaptic Plasticity Dynamics for Deep Continuous Local Learning (DECOLLE)
• Learned Step Size Quantization
• Neuromorphic Deep Learning Machines
• First-spike based visual categorization using reward-modulated STDP
• S4NN: temporal backpropagation for spiking neural networks with one spike per neuron
• A Biologically Plausible Supervised Learning Method for Spiking Neural Networks Using the Symmetric STDP Rule
• CirCNN: Accelerating and Compressing Deep Neural Networks Using Block-CirculantWeight Matrices
• Opportunities in Quantum Reservoir Computing and Extreme Learning Machines
• Survey of Machine Learning Accelerators
• Deep learning incorporating biologically-inspired neural dynamics
• Surrogate Gradient Learning in Spiking Neural Networks
• Event-Based Backpropagation can compute Exact Gradients for Spiking Neural Networks
• Training Spiking Deep Networks for Neuromorphic Hardware
• Theory and Tools for the Conversion of Analog to Spiking Convolutional Neural Networks
• Factorization tricks for LSTM networks
• Whetstone: A Method for Training Deep Artificial Neural Networks for Binary Communication
• Silicon Photonic Architecture for Training Deep Neural Networks with Direct Feedback Alignment
• AI Accelerator Survey and Trends
• Scaling Up Silicon Photonic-based Accelerators: Challenges and Opportunities
• BS4NN: Binarized Spiking Neural Networks with Temporal Coding and Learning
• Applications and Techniques for Fast Machine Learning in Science
• Deep neural networks are robust to weight binarization and other non-linear distortions
• Spiking Deep Residual Network
• Soliton crystal Kerr microcombs for high-speed, scalable optical neural networks at 10 GigaOPs/s
• Crossbar-aware neural network pruning
• Long short-term memory and learning-to-learn in networks of spiking neurons
• Pruning and Quantization for Deep Neural Network Acceleration: A Survey
• DFTerNet: Towards 2-bit Dynamic Fusion Networks for Accurate Human Activity Recognition
• Leveraging Chaos for Wave-Based Analog Computation: Demonstration with Indoor Wireless Communication Signals
• Efficient Processing of Deep Neural Networks: A Tutorial and Survey
• Unsupervised Visual Feature Learning with Spike-timing-dependent Plasticity: How Far are we from Traditional Feature Learning Approaches?
• Early Seizure Detection with an Energy-Efficient Convolutional Neural Network on an Implantable Microcontroller
• Photonic convolutional neural networks using integrated diffractive optics
• Fast and Efficient Asynchronous Neural Computation with Adapting Spiking Neural Networks
• Metabolic constraints on synaptic learning and memory
• Discovering Low-Precision Networks Close to Full-Precision Networks for Efficient Embedded Inference
• Delta Networks for Optimized Recurrent Network Computation
• The Backpropagation Algorithm Implemented on Spiking Neuromorphic Hardware
• Accelerated physical emulation of Bayesian inference in spiking neural networks
• RANC: Reconfigurable Architecture for Neuromorphic Computing
• SC-DCNN: Highly-Scalable Deep Convolutional Neural Network using Stochastic Computing
• RMP-SNN: Residual Membrane Potential Neuron for Enabling Deeper High-Accuracy and Low-Latency Spiking Neural Network
• When Neurons Fail
• Spiking neural networks trained with backpropagation for low power neuromorphic implementation of voice activity detection
• Keys to Accurate Feature Extraction Using Residual Spiking Neural Networks
• Training Deep Spiking Neural Networks
• Quantum-limited stochastic optical neural networks operating at a few quanta per activation
• DeepFire2: A Convolutional Spiking Neural Network Accelerator on FPGAs
• Reservoirs learn to learn
• Integration of Leaky-Integrate-and-Fire-Neurons in Deep Learning Architectures
• Pattern representation and recognition with accelerated analog neuromorphic systems
• C-LSTM: Enabling Efficient LSTM using Structured Compression Techniques on FPGAs
• The wisdom of networks: A general adaptation and learning mechanism of complex systems: The network core triggers fast responses to known stimuli; innovations require the slow network periphery and are encoded by core-remodeling
• Spiking neural networks trained via proxy
• Dendritic-Inspired Processing Enables Bio-Plausible STDP in Compound Binary Synapses
• Efficient Computation in Adaptive Artificial Spiking Neural Networks
• Differentiable graph-structured models for inverse design of lattice materials
• A Tandem Learning Rule for Effective Training and Rapid Inference of Deep Spiking Neural Networks
• Self-consistent formulations for stochastic nonlinear neuronal dynamics
• Evolutionary Spiking Neural Networks: A Survey
• SpikeGrad: An ANN-equivalent Computation Model for Implementing Backpropagation with Spikes
• Comprehensive SNN Compression Using ADMM Optimization and Activity Regularization
• HyPar: Towards Hybrid Parallelism for Deep Learning Accelerator Array
• Sigma Delta Quantized Networks
• Structured Convolution Matrices for Energy-efficient Deep learning
• Visual Odometry with Neuromorphic Resonator Networks
• Deep Learning with the Random Neural Network and its Applications
• Bayesian Photonic Accelerators for Energy Efficient and Noise Robust Neural Processing
• Fractional-order spike-timing-dependent gradient descent for multi-layer spiking neural networks
• Low-Power Neuromorphic Hardware for Signal Processing Applications
• Kinetic mechanism for reversible structural transition in MoTe2 induced by excess charge carriers
• A Power-Efficient Binary-Weight Spiking Neural Network Architecture for Real-Time Object Classification
• Scaling Binarized Neural Networks on Reconfigurable Logic
• GXNOR-Net: Training deep neural networks with ternary weights and activations without full-precision memory under a unified discretization framework
• Recognizing Images with at most one Spike per Neuron
• Spike-based computation using classical recurrent neural networks
• Fault-Tolerant Neural Networks from Biological Error Correction Codes
• Can biological quantum networks solve NP-hard problems?
• Photonic single perceptron at Giga-OP/s speeds with Kerr microcombs for scalable optical neural networks
• FFT-Based Deep Learning Deployment in Embedded Systems
• Distillation Strategies for Proximal Policy Optimization
• Approximations in Deep Learning
• Sparse Spiking Gradient Descent
• Data-driven effective model shows a liquid-like deep learning
• Towards the Next Generation of Retinal Neuroprosthesis: Visual Computation with Spikes
• Low Precision Policy Distillation with Application to Low-Power, Real-time Sensation-Cognition-Action Loop with Neuromorphic Computing
• Large-scale image analysis using docker sandboxing
• Multi-Tones' Phase Coding (MTPC) of Interaural Time Difference by Spiking Neural Network
• Neural Network Training on In-memory-computing Hardware with Radix-4 Gradients
• Artificial neurons based on antiferromagnetic auto-oscillators as a platform for neuromorphic computing
• Computability and Complexity of Unconventional Computing Devices
• D-SELD: Dataset-Scalable Exemplar LCA-Decoder
• Hardware-friendly Neural Network Architecture for Neuromorphic Computing
• Floating-Point Multiplication Using Neuromorphic Computing
• Recurrent and Spiking Modeling of Sparse Surgical Kinematics
• RRAM based neuromorphic algorithms
• Energy Saving Additive Neural Network
• Lincoln AI Computing Survey (LAICS) and Trends
• Layer-wise synapse optimization for implementing neural networks on general neuromorphic architectures
• Arch-Net: Model Distillation for Architecture Agnostic Model Deployment
• A Memristor based Unsupervised Neuromorphic System Towards Fast and Energy-Efficient GAN
• Towards Accurate and High-Speed Spiking Neuromorphic Systems with Data Quantization-Aware Deep Networks
• Reconstruction of Natural Visual Scenes from Neural Spikes with Deep Neural Networks
• Event-based Signal Processing for Radioisotope Identification
• Neural and Synaptic Array Transceiver: A Brain-Inspired Computing Framework for Embedded Learning
• Shenjing: A low power reconfigurable neuromorphic accelerator with partial-sum and spike networks-on-chip
• Robustness from structure: Inference with hierarchical spiking networks on analog neuromorphic hardware
• Single photonic perceptron based on a soliton crystal Kerr microcomb for high-speed, scalable, optical neural networks
• AIDA: Associative DNN Inference Accelerator
• Harnessing Slow Dynamics in Neuromorphic Computation
• DTNN: Energy-efficient Inference with Dendrite Tree Inspired Neural Networks for Edge Vision Applications
• Enabling Bio-Plausible Multi-level STDP using CMOS Neurons with Dendrites and Bistable RRAMs
• 2022 Roadmap on Neuromorphic Computing and Engineering
• Spiking label propagation for community detection
• Spiking Linear Dynamical Systems on Neuromorphic Hardware for Low-Power Brain-Machine Interfaces
• On-Chip Error-triggered Learning of Multi-layer Memristive Spiking Neural Networks
• Implementation of Optical Deep Neural Networks using the Fabry-Perot Interferometer
• ItNet: iterative neural networks with small graphs for accurate, efficient and anytime semantic segmentation
• Threshold Modulation for Online Test-Time Adaptation of Spiking Neural Networks