1
|
Grimaldi A, Boutin V, Ieng SH, Benosman R, Perrinet LU. A robust event-driven approach to always-on object recognition. Neural Netw 2024; 178:106415. [PMID: 38852508 DOI: 10.1016/j.neunet.2024.106415] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 06/14/2023] [Revised: 04/05/2024] [Accepted: 05/29/2024] [Indexed: 06/11/2024]
Abstract
We propose a neuromimetic architecture capable of always-on pattern recognition, i.e. at any time during processing. To achieve this, we have extended an existing event-based algorithm (Lagorce et al., 2017), which introduced novel spatio-temporal features as a Hierarchy Of Time-Surfaces (HOTS). Built from asynchronous events captured by a neuromorphic camera, these time surfaces allow to encode the local dynamics of a visual scene and to create an efficient event-based pattern recognition architecture. Inspired by neuroscience, we have extended this method to improve its performance. First, we add a homeostatic gain control on the activity of neurons to improve the learning of spatio-temporal patterns (Grimaldi et al., 2021). We also provide a new mathematical formalism that allows an analogy to be drawn between the HOTS algorithm and Spiking Neural Networks (SNN). Following this analogy, we transform the offline pattern categorization method into an online and event-driven layer. This classifier uses the spiking output of the network to define new time surfaces and we then perform the online classification with a neuromimetic implementation of a multinomial logistic regression. These improvements not only consistently increase the performance of the network, but also bring this event-driven pattern recognition algorithm fully online. The results have been validated on different datasets: Poker-DVS (Serrano-Gotarredona and Linares-Barranco, 2015), N-MNIST (Orchard, Jayawant et al., 2015) and DVS Gesture (Amir et al., 2017). This demonstrates the efficiency of this bio-realistic SNN for ultra-fast object recognition through an event-by-event categorization process.
Collapse
Affiliation(s)
- Antoine Grimaldi
- Aix-Marseille Universit, Institut de Neurosciences de la Timone, CNRS, Marseille, France.
| | - Victor Boutin
- Carney Institute for Brain Science, Brown University, Providence, RI, United States; Artificial and Natural Intelligence Toulouse Institute, Université de Toulouse, Toulouse, France.
| | - Sio-Hoi Ieng
- Institut de la Vision, Sorbonne Université, CNRS, Paris, France.
| | - Ryad Benosman
- Robotics Institute, Carnegie Mellon University, Pittsburg, PA, United States.
| | - Laurent U Perrinet
- Aix-Marseille Universit, Institut de Neurosciences de la Timone, CNRS, Marseille, France.
| |
Collapse
|
2
|
Dorzhigulov A, Saxena V. Spiking CMOS-NVM mixed-signal neuromorphic ConvNet with circuit- and training-optimized temporal subsampling. Front Neurosci 2023; 17:1177592. [PMID: 37534034 PMCID: PMC10390782 DOI: 10.3389/fnins.2023.1177592] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 03/01/2023] [Accepted: 06/26/2023] [Indexed: 08/04/2023] Open
Abstract
We increasingly rely on deep learning algorithms to process colossal amount of unstructured visual data. Commonly, these deep learning algorithms are deployed as software models on digital hardware, predominantly in data centers. Intrinsic high energy consumption of Cloud-based deployment of deep neural networks (DNNs) inspired researchers to look for alternatives, resulting in a high interest in Spiking Neural Networks (SNNs) and dedicated mixed-signal neuromorphic hardware. As a result, there is an emerging challenge to transfer DNN architecture functionality to energy-efficient spiking non-volatile memory (NVM)-based hardware with minimal loss in the accuracy of visual data processing. Convolutional Neural Network (CNN) is the staple choice of DNN for visual data processing. However, the lack of analog-friendly spiking implementations and alternatives for some core CNN functions, such as MaxPool, hinders the conversion of CNNs into the spike domain, thus hampering neuromorphic hardware development. To address this gap, in this work, we propose MaxPool with temporal multiplexing for Spiking CNNs (SCNNs), which is amenable for implementation in mixed-signal circuits. In this work, we leverage the temporal dynamics of internal membrane potential of Integrate & Fire neurons to enable MaxPool decision-making in the spiking domain. The proposed MaxPool models are implemented and tested within the SCNN architecture using a modified version of the aihwkit framework, a PyTorch-based toolkit for modeling and simulating hardware-based neural networks. The proposed spiking MaxPool scheme can decide even before the complete spatiotemporal input is applied, thus selectively trading off latency with accuracy. It is observed that by allocating just 10% of the spatiotemporal input window for a pooling decision, the proposed spiking MaxPool achieves up to 61.74% accuracy with a 2-bit weight resolution in the CIFAR10 dataset classification task after training with back propagation, with only about 1% performance drop compared to 62.78% accuracy of the 100% spatiotemporal window case with the 2-bit weight resolution to reflect foundry-integrated ReRAM limitations. In addition, we propose the realization of one of the proposed spiking MaxPool techniques in an NVM crossbar array along with periphery circuits designed in a 130nm CMOS technology. The energy-efficiency estimation results show competitive performance compared to recent neuromorphic chip designs.
Collapse
|
3
|
Galinsky VL, Frank LR. Critically synchronized brain waves form an effective, robust and flexible basis for human memory and learning. Sci Rep 2023; 13:4343. [PMID: 36928606 PMCID: PMC10020450 DOI: 10.1038/s41598-023-31365-6] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [MESH Headings] [Grants] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 01/11/2023] [Accepted: 03/10/2023] [Indexed: 03/18/2023] Open
Abstract
The effectiveness, robustness, and flexibility of memory and learning constitute the very essence of human natural intelligence, cognition, and consciousness. However, currently accepted views on these subjects have, to date, been put forth without any basis on a true physical theory of how the brain communicates internally via its electrical signals. This lack of a solid theoretical framework has implications not only for our understanding of how the brain works, but also for wide range of computational models developed from the standard orthodox view of brain neuronal organization and brain network derived functioning based on the Hodgkin-Huxley ad-hoc circuit analogies that have produced a multitude of Artificial, Recurrent, Convolution, Spiking, etc., Neural Networks (ARCSe NNs) that have in turn led to the standard algorithms that form the basis of artificial intelligence (AI) and machine learning (ML) methods. Our hypothesis, based upon our recently developed physical model of weakly evanescent brain wave propagation (WETCOW) is that, contrary to the current orthodox model that brain neurons just integrate and fire under accompaniment of slow leaking, they can instead perform much more sophisticated tasks of efficient coherent synchronization/desynchronization guided by the collective influence of propagating nonlinear near critical brain waves, the waves that currently assumed to be nothing but inconsequential subthreshold noise. In this paper we highlight the learning and memory capabilities of our WETCOW framework and then apply it to the specific application of AI/ML and Neural Networks. We demonstrate that the learning inspired by these critically synchronized brain waves is shallow, yet its timing and accuracy outperforms deep ARCSe counterparts on standard test datasets. These results have implications for both our understanding of brain function and for the wide range of AI/ML applications.
Collapse
Affiliation(s)
- Vitaly L Galinsky
- Center for Scientific Computation in Imaging, University of California at San Diego, La Jolla, CA, 92037-0854, USA.
| | - Lawrence R Frank
- Center for Scientific Computation in Imaging, University of California at San Diego, La Jolla, CA, 92037-0854, USA
- Center for Functional MRI, University of California at San Diego, La Jolla, CA, 92037-0677, USA
| |
Collapse
|
4
|
Effective Conversion of a Convolutional Neural Network into a Spiking Neural Network for Image Recognition Tasks. APPLIED SCIENCES-BASEL 2022. [DOI: 10.3390/app12115749] [Citation(s) in RCA: 2] [Impact Index Per Article: 1.0] [Reference Citation Analysis] [Abstract] [Track Full Text] [Subscribe] [Scholar Register] [Indexed: 02/04/2023]
Abstract
Due to energy efficiency, spiking neural networks (SNNs) have gradually been considered as an alternative to convolutional neural networks (CNNs) in various machine learning tasks. In image recognition tasks, leveraging the superior capability of CNNs, the CNN–SNN conversion is considered one of the most successful approaches to training SNNs. However, previous works assume a rather long inference time period called inference latency to be allowed, while having a trade-off between inference latency and accuracy. One of the main reasons for this phenomenon stems from the difficulty in determining proper a firing threshold for spiking neurons. The threshold determination procedure is called a threshold balancing technique in the CNN–SNN conversion approach. This paper proposes a CNN–SNN conversion method with a new threshold balancing technique that obtains converted SNN models with good accuracy even with low latency. The proposed method organizes the SNN models with soft-reset IF spiking neurons. The threshold balancing technique estimates the thresholds for spiking neurons based on the maximum input current in a layerwise and channelwise manner. The experiment results have shown that our converted SNN models attain even higher accuracy than the corresponding trained CNN model for the MNIST dataset with low latency. In addition, for the Fashion-MNIST and CIFAR-10 datasets, our converted SNNs have shown less conversion loss than other methods in low latencies. The proposed method can be beneficial in deploying efficient SNN models for recognition tasks on resource-limited systems because the inference latency is strongly associated with energy consumption.
Collapse
|
5
|
Faghihi F, Alashwal H, Moustafa AA. A Synaptic Pruning-Based Spiking Neural Network for Hand-Written Digits Classification. Front Artif Intell 2022; 5:680165. [PMID: 35280233 PMCID: PMC8908262 DOI: 10.3389/frai.2022.680165] [Citation(s) in RCA: 3] [Impact Index Per Article: 1.5] [Reference Citation Analysis] [Abstract] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 03/13/2021] [Accepted: 01/14/2022] [Indexed: 12/21/2022] Open
Abstract
A spiking neural network model inspired by synaptic pruning is developed and trained to extract features of hand-written digits. The network is composed of three spiking neural layers and one output neuron whose firing rate is used for classification. The model detects and collects the geometric features of the images from the Modified National Institute of Standards and Technology database (MNIST). In this work, a novel learning rule is developed to train the network to detect features of different digit classes. For this purpose, randomly initialized synaptic weights between the first and second layers are updated using average firing rates of pre- and postsynaptic neurons. Then, using a neuroscience-inspired mechanism named, “synaptic pruning” and its predefined threshold values, some of the synapses are deleted. Hence, these sparse matrices named, “information channels” are constructed so that they show highly specific patterns for each digit class as connection matrices between the first and second layers. The “information channels” are used in the test phase to assign a digit class to each test image. In addition, the role of feed-back inhibition as well as the connectivity rates of the second and third neural layers are studied. Similar to the abilities of the humans to learn from small training trials, the developed spiking neural network needs a very small dataset for training, compared to the conventional deep learning methods that have shown a very good performance on the MNIST dataset. This work introduces a new class of brain-inspired spiking neural networks to extract the features of complex data images.
Collapse
Affiliation(s)
| | - Hany Alashwal
- College of Information Technology, United Arab Emirates University, Al Ain, United Arab Emirates
- *Correspondence: Hany Alashwal
| | - Ahmed A. Moustafa
- School of Psychology, Faculty of Society and Design, Bond University, Gold Coast, QLD, Australia
- Department of Human Anatomy and Physiology, The Faculty of Health Sciences, University of Johannesburg, Johannesburg, South Africa
| |
Collapse
|
6
|
Chen Y, Mai Y, Feng R, Xiao J. An adaptive threshold mechanism for accurate and efficient deep spiking convolutional neural networks. Neurocomputing 2022. [DOI: 10.1016/j.neucom.2021.10.080] [Citation(s) in RCA: 2] [Impact Index Per Article: 1.0] [Reference Citation Analysis] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Indexed: 11/16/2022]
|
7
|
Srinivasan G, Roy K. BlocTrain: Block-Wise Conditional Training and Inference for Efficient Spike-Based Deep Learning. Front Neurosci 2021; 15:603433. [PMID: 34776834 PMCID: PMC8586528 DOI: 10.3389/fnins.2021.603433] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 09/06/2020] [Accepted: 07/23/2021] [Indexed: 12/04/2022] Open
Abstract
Spiking neural networks (SNNs), with their inherent capability to learn sparse spike-based input representations over time, offer a promising solution for enabling the next generation of intelligent autonomous systems. Nevertheless, end-to-end training of deep SNNs is both compute- and memory-intensive because of the need to backpropagate error gradients through time. We propose BlocTrain, which is a scalable and complexity-aware incremental algorithm for memory-efficient training of deep SNNs. We divide a deep SNN into blocks, where each block consists of few convolutional layers followed by a classifier. We train the blocks sequentially using local errors from the classifier. Once a given block is trained, our algorithm dynamically figures out easy vs. hard classes using the class-wise accuracy, and trains the deeper block only on the hard class inputs. In addition, we also incorporate a hard class detector (HCD) per block that is used during inference to exit early for the easy class inputs and activate the deeper blocks only for the hard class inputs. We trained ResNet-9 SNN divided into three blocks, using BlocTrain, on CIFAR-10 and obtained 86.4% accuracy, which is achieved with up to 2.95× lower memory requirement during the course of training, and 1.89× compute efficiency per inference (due to early exit strategy) with 1.45× memory overhead (primarily due to classifier weights) compared to end-to-end network. We also trained ResNet-11, divided into four blocks, on CIFAR-100 and obtained 58.21% accuracy, which is one of the first reported accuracy for SNN trained entirely with spike-based backpropagation on CIFAR-100.
Collapse
Affiliation(s)
- Gopalakrishnan Srinivasan
- Department of Electrical and Computer Engineering, Purdue University, West Lafayette, IN, United States
| | - Kaushik Roy
- Department of Electrical and Computer Engineering, Purdue University, West Lafayette, IN, United States
| |
Collapse
|
8
|
A Low-Cost Hardware-Friendly Spiking Neural Network Based on Binary MRAM Synapses, Accelerated Using In-Memory Computing. ELECTRONICS 2021. [DOI: 10.3390/electronics10192441] [Citation(s) in RCA: 1] [Impact Index Per Article: 0.3] [Reference Citation Analysis] [Abstract] [Track Full Text] [Subscribe] [Scholar Register] [Indexed: 11/16/2022]
Abstract
In recent years, the scaling down that Moore’s Law relies on has been gradually slowing down, and the traditional von Neumann architecture has been limiting the improvement of computing power. Thus, neuromorphic in-memory computing hardware has been proposed and is becoming a promising alternative. However, there is still a long way to make it possible, and one of the problems is to provide an efficient, reliable, and achievable neural network for hardware implementation. In this paper, we proposed a two-layer fully connected spiking neural network based on binary MRAM (Magneto-resistive Random Access Memory) synapses with low hardware cost. First, the network used an array of multiple binary MRAM cells to store multi-bit fixed-point weight values. This helps to simplify the read/write circuit. Second, we used different kinds of spike encoders that ensure the sparsity of input spikes, to reduce the complexity of peripheral circuits, such as sense amplifiers. Third, we designed a single-step learning rule, which fit well with the fixed-point binary weights. Fourth, we replaced the traditional exponential Leak-Integrate-Fire (LIF) neuron model to avoid the massive cost of exponential circuits. The simulation results showed that, compared to other similar works, our SNN with 1184 neurons and 313,600 synapses achieved an accuracy of up to 90.6% in the MNIST recognition task with full-resolution (28 × 28) and full-bit-depth (8-bit) images. In the case of low-resolution (16 × 16) and black-white (1-bit) images, the smaller version of our network with 384 neurons and 32,768 synapses still maintained an accuracy of about 77%, extending its application to ultra-low-cost situations. Both versions need less than 30,000 samples to reach convergence, which is a >50% reduction compared to other similar networks. As for robustness, it is immune to the fluctuation of MRAM cell resistance.
Collapse
|
9
|
A Review of Algorithms and Hardware Implementations for Spiking Neural Networks. JOURNAL OF LOW POWER ELECTRONICS AND APPLICATIONS 2021. [DOI: 10.3390/jlpea11020023] [Citation(s) in RCA: 7] [Impact Index Per Article: 2.3] [Reference Citation Analysis] [Abstract] [Track Full Text] [Subscribe] [Scholar Register] [Indexed: 11/17/2022]
Abstract
Deep Learning (DL) has contributed to the success of many applications in recent years. The applications range from simple ones such as recognizing tiny images or simple speech patterns to ones with a high level of complexity such as playing the game of Go. However, this superior performance comes at a high computational cost, which made porting DL applications to conventional hardware platforms a challenging task. Many approaches have been investigated, and Spiking Neural Network (SNN) is one of the promising candidates. SNN is the third generation of Artificial Neural Networks (ANNs), where each neuron in the network uses discrete spikes to communicate in an event-based manner. SNNs have the potential advantage of achieving better energy efficiency than their ANN counterparts. While generally there will be a loss of accuracy on SNN models, new algorithms have helped to close the accuracy gap. For hardware implementations, SNNs have attracted much attention in the neuromorphic hardware research community. In this work, we review the basic background of SNNs, the current state and challenges of the training algorithms for SNNs and the current implementations of SNNs on various hardware platforms.
Collapse
|
10
|
Debat G, Chauhan T, Cottereau BR, Masquelier T, Paindavoine M, Baures R. Event-Based Trajectory Prediction Using Spiking Neural Networks. Front Comput Neurosci 2021; 15:658764. [PMID: 34108870 PMCID: PMC8180888 DOI: 10.3389/fncom.2021.658764] [Citation(s) in RCA: 2] [Impact Index Per Article: 0.7] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 01/26/2021] [Accepted: 04/27/2021] [Indexed: 11/13/2022] Open
Abstract
In recent years, event-based sensors have been combined with spiking neural networks (SNNs) to create a new generation of bio-inspired artificial vision systems. These systems can process spatio-temporal data in real time, and are highly energy efficient. In this study, we used a new hybrid event-based camera in conjunction with a multi-layer spiking neural network trained with a spike-timing-dependent plasticity learning rule. We showed that neurons learn from repeated and correlated spatio-temporal patterns in an unsupervised way and become selective to motion features, such as direction and speed. This motion selectivity can then be used to predict ball trajectory by adding a simple read-out layer composed of polynomial regressions, and trained in a supervised manner. Hence, we show that a SNN receiving inputs from an event-based sensor can extract relevant spatio-temporal patterns to process and predict ball trajectories.
Collapse
Affiliation(s)
- Guillaume Debat
- CERCO UMR 5549, CNRS-Université Toulouse 3, Toulouse, France
| | - Tushar Chauhan
- CERCO UMR 5549, CNRS-Université Toulouse 3, Toulouse, France
| | | | | | - Michel Paindavoine
- Laboratory for Research on Learning and Development (LEAD), University of Burgundy, CNRS UMR, Dijon, France
| | - Robin Baures
- CERCO UMR 5549, CNRS-Université Toulouse 3, Toulouse, France
| |
Collapse
|
11
|
Iyer LR, Chua Y, Li H. Is Neuromorphic MNIST Neuromorphic? Analyzing the Discriminative Power of Neuromorphic Datasets in the Time Domain. Front Neurosci 2021; 15:608567. [PMID: 33841072 PMCID: PMC8027306 DOI: 10.3389/fnins.2021.608567] [Citation(s) in RCA: 6] [Impact Index Per Article: 2.0] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 09/21/2020] [Accepted: 03/01/2021] [Indexed: 11/26/2022] Open
Abstract
A major characteristic of spiking neural networks (SNNs) over conventional artificial neural networks (ANNs) is their ability to spike, enabling them to use spike timing for coding and efficient computing. In this paper, we assess if neuromorphic datasets recorded from static images are able to evaluate the ability of SNNs to use spike timings in their calculations. We have analyzed N-MNIST, N-Caltech101 and DvsGesture along these lines, but focus our study on N-MNIST. First we evaluate if additional information is encoded in the time domain in a neuromorphic dataset. We show that an ANN trained with backpropagation on frame-based versions of N-MNIST and N-Caltech101 images achieve 99.23 and 78.01% accuracy. These are comparable to the state of the art-showing that an algorithm that purely works on spatial data can classify these datasets. Second we compare N-MNIST and DvsGesture on two STDP algorithms, RD-STDP, that can classify only spatial data, and STDP-tempotron that classifies spatiotemporal data. We demonstrate that RD-STDP performs very well on N-MNIST, while STDP-tempotron performs better on DvsGesture. Since DvsGesture has a temporal dimension, it requires STDP-tempotron, while N-MNIST can be adequately classified by an algorithm that works on spatial data alone. This shows that precise spike timings are not important in N-MNIST. N-MNIST does not, therefore, highlight the ability of SNNs to classify temporal data. The conclusions of this paper open the question-what dataset can evaluate SNN ability to classify temporal data?
Collapse
Affiliation(s)
- Laxmi R. Iyer
- Neuromorphic Computing, Institute of Infocomms Research, A*Star, Singapore, Singapore
| | - Yansong Chua
- Neuromorphic Computing, Institute of Infocomms Research, A*Star, Singapore, Singapore
| | - Haizhou Li
- Neuromorphic Computing, Institute of Infocomms Research, A*Star, Singapore, Singapore
- Huawei Technologies Co., Ltd., Shenzhen, China
| |
Collapse
|
12
|
Rank order coding based spiking convolutional neural network architecture with energy-efficient membrane voltage updates. Neurocomputing 2020. [DOI: 10.1016/j.neucom.2020.05.031] [Citation(s) in RCA: 7] [Impact Index Per Article: 1.8] [Reference Citation Analysis] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Indexed: 11/21/2022]
|
13
|
Efficient and hardware-friendly methods to implement competitive learning for spiking neural networks. Neural Comput Appl 2020. [DOI: 10.1007/s00521-020-04755-4] [Citation(s) in RCA: 2] [Impact Index Per Article: 0.5] [Reference Citation Analysis] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Indexed: 10/25/2022]
|
14
|
Abderrahmane N, Lemaire E, Miramond B. Design Space Exploration of Hardware Spiking Neurons for Embedded Artificial Intelligence. Neural Netw 2019; 121:366-386. [PMID: 31593842 DOI: 10.1016/j.neunet.2019.09.024] [Citation(s) in RCA: 16] [Impact Index Per Article: 3.2] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 02/27/2019] [Revised: 09/16/2019] [Accepted: 09/17/2019] [Indexed: 11/30/2022]
Abstract
Machine learning is yielding unprecedented interest in research and industry, due to recent success in many applied contexts such as image classification and object recognition. However, the deployment of these systems requires huge computing capabilities, thus making them unsuitable for embedded systems. To deal with this limitation, many researchers are investigating brain-inspired computing, which would be a perfect alternative to the conventional Von Neumann architecture based computers (CPU/GPU) that meet the requirements for computing performance, but not for energy-efficiency. Therefore, neuromorphic hardware circuits that are adaptable for both parallel and distributed computations need to be designed. In this paper, we focus on Spiking Neural Networks (SNNs) with a comprehensive study of neural coding methods and hardware exploration. In this context, we propose a framework for neuromorphic hardware design space exploration, which allows to define a suitable architecture based on application-specific constraints and starting from a wide variety of possible architectural choices. For this framework, we have developed a behavioral level simulator for neuromorphic hardware architectural exploration named NAXT. Moreover, we propose modified versions of the standard Rate Coding technique to make trade-offs with the Time Coding paradigm, which is characterized by the low number of spikes propagating in the network. Thus, we are able to reduce the number of spikes while keeping the same neuron's model, which results in an SNN with fewer events to process. By doing so, we seek to reduce the amount of power consumed by the hardware. Furthermore, we present three neuromorphic hardware architectures in order to quantitatively study the implementation of SNNs. One of these architectures integrates a novel hybrid structure: a highly-parallel computation core for most solicited layers, and time-multiplexed computation units for deeper layers. These architectures are derived from a novel funnel-like Design Space Exploration framework for neuromorphic hardware.
Collapse
Affiliation(s)
| | - Edgar Lemaire
- Université Côte d'Azur, CNRS, LEAT, France; Thales Research Technology / STI Group / LCHP, Palaiseau, France.
| | | |
Collapse
|
15
|
Mozafari M, Ganjtabesh M, Nowzari-Dalini A, Masquelier T. SpykeTorch: Efficient Simulation of Convolutional Spiking Neural Networks With at Most One Spike per Neuron. Front Neurosci 2019; 13:625. [PMID: 31354403 PMCID: PMC6640212 DOI: 10.3389/fnins.2019.00625] [Citation(s) in RCA: 21] [Impact Index Per Article: 4.2] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 03/01/2019] [Accepted: 05/31/2019] [Indexed: 11/13/2022] Open
Abstract
Application of deep convolutional spiking neural networks (SNNs) to artificial intelligence (AI) tasks has recently gained a lot of interest since SNNs are hardware-friendly and energy-efficient. Unlike the non-spiking counterparts, most of the existing SNN simulation frameworks are not practically efficient enough for large-scale AI tasks. In this paper, we introduce SpykeTorch, an open-source high-speed simulation framework based on PyTorch. This framework simulates convolutional SNNs with at most one spike per neuron and the rank-order encoding scheme. In terms of learning rules, both spike-timing-dependent plasticity (STDP) and reward-modulated STDP (R-STDP) are implemented, but other rules could be implemented easily. Apart from the aforementioned properties, SpykeTorch is highly generic and capable of reproducing the results of various studies. Computations in the proposed framework are tensor-based and totally done by PyTorch functions, which in turn brings the ability of just-in-time optimization for running on CPUs, GPUs, or Multi-GPU platforms.
Collapse
Affiliation(s)
- Milad Mozafari
- Department of Computer Science, School of Mathematics, Statistics, and Computer Science, University of Tehran, Tehran, Iran.,CERCO UMR 5549, CNRS - Université Toulouse 3, Toulouse, France
| | - Mohammad Ganjtabesh
- Department of Computer Science, School of Mathematics, Statistics, and Computer Science, University of Tehran, Tehran, Iran
| | - Abbas Nowzari-Dalini
- Department of Computer Science, School of Mathematics, Statistics, and Computer Science, University of Tehran, Tehran, Iran
| | | |
Collapse
|
16
|
Illing B, Gerstner W, Brea J. Biologically plausible deep learning - But how far can we go with shallow networks? Neural Netw 2019; 118:90-101. [PMID: 31254771 DOI: 10.1016/j.neunet.2019.06.001] [Citation(s) in RCA: 30] [Impact Index Per Article: 6.0] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 02/07/2019] [Revised: 05/29/2019] [Accepted: 06/02/2019] [Indexed: 11/17/2022]
Abstract
Training deep neural networks with the error backpropagation algorithm is considered implausible from a biological perspective. Numerous recent publications suggest elaborate models for biologically plausible variants of deep learning, typically defining success as reaching around 98% test accuracy on the MNIST data set. Here, we investigate how far we can go on digit (MNIST) and object (CIFAR10) classification with biologically plausible, local learning rules in a network with one hidden layer and a single readout layer. The hidden layer weights are either fixed (random or random Gabor filters) or trained with unsupervised methods (Principal/Independent Component Analysis or Sparse Coding) that can be implemented by local learning rules. The readout layer is trained with a supervised, local learning rule. We first implement these models with rate neurons. This comparison reveals, first, that unsupervised learning does not lead to better performance than fixed random projections or Gabor filters for large hidden layers. Second, networks with localized receptive fields perform significantly better than networks with all-to-all connectivity and can reach backpropagation performance on MNIST. We then implement two of the networks - fixed, localized, random & random Gabor filters in the hidden layer - with spiking leaky integrate-and-fire neurons and spike timing dependent plasticity to train the readout layer. These spiking models achieve >98.2% test accuracy on MNIST, which is close to the performance of rate networks with one hidden layer trained with backpropagation. The performance of our shallow network models is comparable to most current biologically plausible models of deep learning. Furthermore, our results with a shallow spiking network provide an important reference and suggest the use of data sets other than MNIST for testing the performance of future models of biologically plausible deep learning.
Collapse
Affiliation(s)
- Bernd Illing
- School of Computer and Communication Science & School of Life Science, EPFL, 1015 Lausanne, Switzerland.
| | - Wulfram Gerstner
- School of Computer and Communication Science & School of Life Science, EPFL, 1015 Lausanne, Switzerland
| | - Johanni Brea
- School of Computer and Communication Science & School of Life Science, EPFL, 1015 Lausanne, Switzerland
| |
Collapse
|
17
|
Srinivasan G, Roy K. ReStoCNet: Residual Stochastic Binary Convolutional Spiking Neural Network for Memory-Efficient Neuromorphic Computing. Front Neurosci 2019; 13:189. [PMID: 30941003 PMCID: PMC6434391 DOI: 10.3389/fnins.2019.00189] [Citation(s) in RCA: 31] [Impact Index Per Article: 6.2] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 11/29/2018] [Accepted: 02/18/2019] [Indexed: 11/13/2022] Open
Abstract
In this work, we propose ReStoCNet, a residual stochastic multilayer convolutional Spiking Neural Network (SNN) composed of binary kernels, to reduce the synaptic memory footprint and enhance the computational efficiency of SNNs for complex pattern recognition tasks. ReStoCNet consists of an input layer followed by stacked convolutional layers for hierarchical input feature extraction, pooling layers for dimensionality reduction, and fully-connected layer for inference. In addition, we introduce residual connections between the stacked convolutional layers to improve the hierarchical feature learning capability of deep SNNs. We propose Spike Timing Dependent Plasticity (STDP) based probabilistic learning algorithm, referred to as Hybrid-STDP (HB-STDP), incorporating Hebbian and anti-Hebbian learning mechanisms, to train the binary kernels forming ReStoCNet in a layer-wise unsupervised manner. We demonstrate the efficacy of ReStoCNet and the presented HB-STDP based unsupervised training methodology on the MNIST and CIFAR-10 datasets. We show that residual connections enable the deeper convolutional layers to self-learn useful high-level input features and mitigate the accuracy loss observed in deep SNNs devoid of residual connections. The proposed ReStoCNet offers >20 × kernel memory compression compared to full-precision (32-bit) SNN while yielding high enough classification accuracy on the chosen pattern recognition tasks.
Collapse
Affiliation(s)
| | - Kaushik Roy
- Department of ECE, Purdue University, West Lafayette, IN, United States
| |
Collapse
|
18
|
Tavanaei A, Ghodrati M, Kheradpisheh SR, Masquelier T, Maida A. Deep learning in spiking neural networks. Neural Netw 2018; 111:47-63. [PMID: 30682710 DOI: 10.1016/j.neunet.2018.12.002] [Citation(s) in RCA: 205] [Impact Index Per Article: 34.2] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 06/07/2018] [Revised: 12/02/2018] [Accepted: 12/03/2018] [Indexed: 12/14/2022]
Abstract
In recent years, deep learning has revolutionized the field of machine learning, for computer vision in particular. In this approach, a deep (multilayer) artificial neural network (ANN) is trained, most often in a supervised manner using backpropagation. Vast amounts of labeled training examples are required, but the resulting classification accuracy is truly impressive, sometimes outperforming humans. Neurons in an ANN are characterized by a single, static, continuous-valued activation. Yet biological neurons use discrete spikes to compute and transmit information, and the spike times, in addition to the spike rates, matter. Spiking neural networks (SNNs) are thus more biologically realistic than ANNs, and are arguably the only viable option if one wants to understand how the brain computes at the neuronal description level. The spikes of biological neurons are sparse in time and space, and event-driven. Combined with bio-plausible local learning rules, this makes it easier to build low-power, neuromorphic hardware for SNNs. However, training deep SNNs remains a challenge. Spiking neurons' transfer function is usually non-differentiable, which prevents using backpropagation. Here we review recent supervised and unsupervised methods to train deep SNNs, and compare them in terms of accuracy and computational cost. The emerging picture is that SNNs still lag behind ANNs in terms of accuracy, but the gap is decreasing, and can even vanish on some tasks, while SNNs typically require many fewer operations and are the better candidates to process spatio-temporal data.
Collapse
Affiliation(s)
- Amirhossein Tavanaei
- School of Computing and Informatics, University of Louisiana at Lafayette, Lafayette, LA 70504, USA.
| | - Masoud Ghodrati
- Department of Physiology, Monash University, Clayton, VIC, Australia
| | - Saeed Reza Kheradpisheh
- Department of Computer Science, Faculty of Mathematical Sciences and Computer, Kharazmi University, Tehran, Iran
| | | | - Anthony Maida
- School of Computing and Informatics, University of Louisiana at Lafayette, Lafayette, LA 70504, USA
| |
Collapse
|
19
|
Abstract
Neuromorphic computing systems are promising alternatives in the fields of pattern recognition, image processing, etc. especially when conventional von Neumann architectures face several bottlenecks. Memristors play vital roles in neuromorphic computing systems and are usually used as synaptic devices. Memristive spiking neural networks (MSNNs) are considered to be more efficient and biologically plausible than other systems due to their spike-based working mechanism. In contrast to previous SNNs with complex architectures, we propose a hardware-friendly architecture and an unsupervised spike-timing dependent plasticity (STDP) learning method for MSNNs in this paper. The architecture, which is friendly to hardware implementation, includes an input layer, a feature learning layer and a voting circuit. To reduce hardware complexity, some constraints are enforced: the proposed architecture has no lateral inhibition and is purely feedforward; it uses the voting circuit as a classifier and does not use additional classifiers; all neurons can generate at most one spike and do not need to consider firing rates and refractory periods; all neurons have the same fixed threshold voltage for classification. The presented unsupervised STDP learning method is time-dependent and uses no homeostatic mechanism. The MNIST dataset is used to demonstrate our proposed architecture and learning method. Simulation results show that our proposed architecture with the learning method achieves a classification accuracy of 94.6%, which outperforms other unsupervised SNNs that use time-based encoding schemes.
Collapse
|
20
|
Pfeiffer M, Pfeil T. Deep Learning With Spiking Neurons: Opportunities and Challenges. Front Neurosci 2018; 12:774. [PMID: 30410432 PMCID: PMC6209684 DOI: 10.3389/fnins.2018.00774] [Citation(s) in RCA: 119] [Impact Index Per Article: 19.8] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 05/22/2018] [Accepted: 10/04/2018] [Indexed: 01/16/2023] Open
Abstract
Spiking neural networks (SNNs) are inspired by information processing in biology, where sparse and asynchronous binary signals are communicated and processed in a massively parallel fashion. SNNs on neuromorphic hardware exhibit favorable properties such as low power consumption, fast inference, and event-driven information processing. This makes them interesting candidates for the efficient implementation of deep neural networks, the method of choice for many machine learning tasks. In this review, we address the opportunities that deep spiking networks offer and investigate in detail the challenges associated with training SNNs in a way that makes them competitive with conventional deep learning, but simultaneously allows for efficient mapping to hardware. A wide range of training methods for SNNs is presented, ranging from the conversion of conventional deep networks into SNNs, constrained training before conversion, spiking variants of backpropagation, and biologically motivated variants of STDP. The goal of our review is to define a categorization of SNN training methods, and summarize their advantages and drawbacks. We further discuss relationships between SNNs and binary networks, which are becoming popular for efficient digital hardware implementation. Neuromorphic hardware platforms have great potential to enable deep spiking networks in real-world applications. We compare the suitability of various neuromorphic systems that have been developed over the past years, and investigate potential use cases. Neuromorphic approaches and conventional machine learning should not be considered simply two solutions to the same classes of problems, instead it is possible to identify and exploit their task-specific advantages. Deep SNNs offer great opportunities to work with new types of event-based sensors, exploit temporal codes and local on-chip learning, and we have so far just scratched the surface of realizing these advantages in practical applications.
Collapse
Affiliation(s)
- Michael Pfeiffer
- Bosch Center for Artificial Intelligence, Robert Bosch GmbH, Renningen, Germany
| | | |
Collapse
|
21
|
Srinivasan G, Panda P, Roy K. SpiLinC: Spiking Liquid-Ensemble Computing for Unsupervised Speech and Image Recognition. Front Neurosci 2018; 12:524. [PMID: 30190670 PMCID: PMC6116788 DOI: 10.3389/fnins.2018.00524] [Citation(s) in RCA: 12] [Impact Index Per Article: 2.0] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 12/06/2017] [Accepted: 07/12/2018] [Indexed: 11/30/2022] Open
Abstract
In this work, we propose a Spiking Neural Network (SNN) consisting of input neurons sparsely connected by plastic synapses to a randomly interlinked liquid, referred to as Liquid-SNN, for unsupervised speech and image recognition. We adapt the strength of the synapses interconnecting the input and liquid using Spike Timing Dependent Plasticity (STDP), which enables the neurons to self-learn a general representation of unique classes of input patterns. The presented unsupervised learning methodology makes it possible to infer the class of a test input directly using the liquid neuronal spiking activity. This is in contrast to standard Liquid State Machines (LSMs) that have fixed synaptic connections between the input and liquid followed by a readout layer (trained in a supervised manner) to extract the liquid states and infer the class of the input patterns. Moreover, the utility of LSMs has primarily been demonstrated for speech recognition. We find that training such LSMs is challenging for complex pattern recognition tasks because of the information loss incurred by using fixed input to liquid synaptic connections. We show that our Liquid-SNN is capable of efficiently recognizing both speech and image patterns by learning the rich temporal information contained in the respective input patterns. However, the need to enlarge the liquid for improving the accuracy introduces scalability challenges and training inefficiencies. We propose SpiLinC that is composed of an ensemble of multiple liquids operating in parallel. We use a “divide and learn” strategy for SpiLinC, where each liquid is trained on a unique segment of the input patterns that causes the neurons to self-learn distinctive input features. SpiLinC effectively recognizes a test pattern by combining the spiking activity of the constituent liquids, each of which identifies characteristic input features. As a result, SpiLinC offers competitive classification accuracy compared to the Liquid-SNN with added sparsity in synaptic connectivity and faster training convergence, both of which lead to improved energy efficiency in neuromorphic hardware implementations. We validate the efficacy of the proposed Liquid-SNN and SpiLinC on the entire digit subset of the TI46 speech corpus and handwritten digits from the MNIST dataset.
Collapse
Affiliation(s)
| | | | - Kaushik Roy
- Department of ECE, Purdue University, West Lafayette, IN, United States
| |
Collapse
|