Open accessJournal Article

# Neural Coding in Spiking Neural Networks: A Comparative Study for Robust Neuromorphic Systems.

04 Mar 2021-Frontiers in Neuroscience (Frontiers Media SA)-Vol. 15, pp 638474-638474
Abstract: Various hypotheses of information representation in brain, referred to as neural codes, have been proposed to explain the information transmission between neurons. Neural coding plays an essential role in enabling the brain-inspired spiking neural networks (SNNs) to perform different tasks. To search for the best coding scheme, we performed an extensive comparative study on the impact and performance of four important neural coding schemes, namely, rate coding, time-to-first spike (TTFS) coding, phase coding, and burst coding. The comparative study was carried out using a biological 2-layer SNN trained with an unsupervised spike-timing-dependent plasticity (STDP) algorithm. Various aspects of network performance were considered, including classification accuracy, processing latency, synaptic operations (SOPs), hardware implementation, network compression efficacy, input and synaptic noise resilience, and synaptic fault tolerance. The classification tasks on Modified National Institute of Standards and Technology (MNIST) and Fashion-MNIST datasets were applied in our study. For hardware implementation, area and power consumption were estimated for these coding schemes, and the network compression efficacy was analyzed using pruning and quantization techniques. Different types of input noise and noise variations in the datasets were considered and applied. Furthermore, the robustness of each coding scheme to the non-ideality-induced synaptic noise and fault in analog neuromorphic systems was studied and compared. Our results show that TTFS coding is the best choice in achieving the highest computational performance with very low hardware implementation overhead. TTFS coding requires 4x/7.5x lower processing latency and 3.5x/6.5x fewer SOPs than rate coding during the training/inference process. Phase coding is the most resilient scheme to input noise. Burst coding offers the highest network compression efficacy and the best overall robustness to hardware non-idealities for both training and inference processes. The study presented in this paper reveals the design space created by the choice of each coding scheme, allowing designers to frame each scheme in terms of its strength and weakness given a designs' constraints and considerations in neuromorphic systems.

##### Citations
More

5 results found

Open accessPosted Content
Abstract: Realizing the potential of mixed-signal neuromorphic processors for ultra-low-power inference and learning requires efficient use of their inhomogeneous analog circuitry as well as sparse, time-based information encoding and processing. Here, we investigate spike-timing-based spatiotemporal receptive fields of output-neurons in the Spatiotemporal Correlator (STC) network, for which we used excitatory-inhibitory balanced disynaptic inputs instead of dedicated axonal or neuronal delays. We present hardware-in-the-loop experiments with a mixed-signal DYNAP-SE neuromorphic processor, in which five-dimensional receptive fields of hardware neurons were mapped by randomly sampling input spike-patterns from a uniform distribution. We find that, when the balanced disynaptic elements are randomly programmed, some of the neurons display distinct receptive fields. Furthermore, we demonstrate how a neuron was tuned to detect a particular spatiotemporal feature, to which it initially was non-selective, by activating a different subset of the inhomogeneous analog synaptic circuits. The energy dissipation of the balanced synaptic elements is one order of magnitude lower per lateral connection (0.65 nJ vs 9.3 nJ per spike) than former delay-based neuromorphic hardware implementations. Thus, we show how the inhomogeneous synaptic circuits could be utilized for resource-efficient implementation of STC network layers, in a way that enables synapse-address reprogramming as a discrete mechanism for feature tuning.

Topics: , Receptive field (51%)

Open accessProceedings Article
Nikhil Garg1, Ismael Balafrej1, Yann Beilliard1, Dominique Drouin1  +2 moreInstitutions (1)
Abstract: Surface electromyogram (sEMG) signals result from muscle movement and hence they are an ideal candidate for benchmarking event-driven sensing and computing. We propose a simple yet novel approach for optimizing the spike encoding algorithm's hyper-parameters inspired by the readout layer concept in reservoir computing. Using a simple machine learning algorithm after spike encoding, we report performance higher than the state-of-the-art spiking neural networks on two open-source datasets for hand gesture recognition. The spike encoded data is processed through a spiking reservoir with a biologically inspired topology and neuron model. When trained with the unsupervised activity regulation CRITICAL algorithm to operate at the edge of chaos, the reservoir yields better performance than state-of-the-art convolutional neural networks. The reservoir performance with regulated activity was found to be 89.72% for the Roshambo EMG dataset and 70.6% for the EMG subset of sensor fusion dataset. Therefore, the biologically-inspired computing paradigm, which is known for being power efficient, also proves to have a great potential when compared with conventional AI algorithms.

Proceedings Article
01 Oct 2021-
Abstract: Image classification is always an interesting problem due to its practical applications in real life. With a capability of self-learning features, modern Convolution Neural Network (CNN) models can achieve high accuracy on large and complex benchmark datasets. However, due to their high computation costs, the CNN models experience energy consumption problems during training and implementation of the hardware which limits their utilisation in mobile and embedded applications. Recently, the Spiking Neural Network (SNN) has been proposed to overcome drawbacks of the CNN models. Like the biological nervous system, the SNN’s neurons communicate with each other by sending spike trains. A neuron is only calculated when a new input spike arrives. As a result, it turns the networks into an energy-saving mode which is suitable for implementation on hardware devices. To avoid the difficulty of the SNN direct training, an indirect training approach is proposed in this work. A proposed CNN model is firstly trained with the RMSprop algorithm then the optimised weights and bias are mapped to the SNN model converted from the proposed CNN model. Experimental results confirm that our model achieves the best accuracy of 93.5% when compared to state-of-the-art SNN approaches on the Fashion- MNIST dataset.

Topics: , MNIST database (54%),

Open accessJournal Article
13 Nov 2021-IEEE Access
Abstract: This paper proposes an in-memory binary spiking neural network (BSNN) based on spin-transfer-torque magnetoresistive RAM (STT-MRAM). We propose residual BSNN learning using a surrogate gradient that shortens the time steps in the BSNN while maintaining sufficient accuracy. At the circuit level, presynaptic spikes are fed to memory units through differential bit lines (BLs), while binarized weights are stored in a subarray of nonvolatile STT-MRAM. When the common inputs are fed through BLs, vector-to-matrix multiplication can be performed in a single memory sensing phase, hence achieving massive parallelism with low power and low latency. We further introduce the concept of a dynamic threshold to reduce the implementation complexity of synapses and neuron circuitry. This adjustable threshold also permits a nonlinear batch normalization (BN) function to be incorporated into the integrate-and-fire (IF) neuron circuit. The circuitry greatly improves the overall performance and enables high regularity in circuit layouts. Our proposed netlist circuits are built on a 65-nm CMOS with a fitted magnetic tunnel junction (MTJ) model for performance evaluation. The hardware/software co-simulation results indicate that the proposed design can deliver a performance of 176.6 TOPS/W for an in-memory computing (IMC) subarray size of $1\times 288$ . The classification accuracy reaches 97.92% (83.85%) on the MNIST (CIFAR-10) dataset. The impacts of the device non-idealities and process variations are also thoroughly covered in the analysis.

Topics: , , Netlist (51%)

Open accessProceedings Article
Nikhil Garg1, Ismael Balafrej1, Yann Beilliard1, Dominique Drouin2  +2 moreInstitutions (3)
27 Jul 2021-
Abstract: Surface electromyogram (sEMG) signals result from muscle movement and hence they are an ideal candidate for benchmarking event-driven sensing and computing. We propose a simple yet novel approach for optimizing the spike encoding algorithm’s hyper-parameters inspired by the readout layer concept in reservoir computing. Using a simple machine learning algorithm after spike encoding, we report performance higher than the state-of-the-art spiking neural networks on two open-source datasets for hand gesture recognition. The spike encoded data is processed through a spiking reservoir with a biologically inspired topology and neuron model. When trained with the unsupervised activity regulation CRITICAL algorithm to operate at the edge of chaos, the reservoir yields better performance than state-of-the-art convolutional neural networks. The reservoir performance with regulated activity was found to be 89.72% for the Roshambo EMG dataset and 70.6% for the EMG subset of sensor fusion dataset. Therefore, the biologically-inspired computing paradigm, which is known for being power efficient, also proves to have a great potential when compared with conventional AI algorithms.

##### References
More

59 results found

Journal ArticleDOI: 10.1109/5.726791
Yann LeCun1, Léon Bottou2, Léon Bottou3, Yoshua Bengio4  +3 moreInstitutions (5)
01 Jan 1998-
Abstract: Multilayer neural networks trained with the back-propagation algorithm constitute the best example of a successful gradient based learning technique. Given an appropriate network architecture, gradient-based learning algorithms can be used to synthesize a complex decision surface that can classify high-dimensional patterns, such as handwritten characters, with minimal preprocessing. This paper reviews various methods applied to handwritten character recognition and compares them on a standard handwritten digit recognition task. Convolutional neural networks, which are specifically designed to deal with the variability of 2D shapes, are shown to outperform all other techniques. Real-life document recognition systems are composed of multiple modules including field extraction, segmentation recognition, and language modeling. A new learning paradigm, called graph transformer networks (GTN), allows such multimodule systems to be trained globally using gradient-based methods so as to minimize an overall performance measure. Two systems for online handwriting recognition are described. Experiments demonstrate the advantage of global training, and the flexibility of graph transformer networks. A graph transformer network for reading a bank cheque is also described. It uses convolutional neural network character recognizers combined with global training techniques to provide record accuracy on business and personal cheques. It is deployed commercially and reads several million cheques per day.

34,930 Citations

Open accessJournal ArticleDOI: 10.1145/3065386
Abstract: We trained a large, deep convolutional neural network to classify the 1.2 million high-resolution images in the ImageNet LSVRC-2010 contest into the 1000 different classes. On the test data, we achieved top-1 and top-5 error rates of 37.5% and 17.0%, respectively, which is considerably better than the previous state-of-the-art. The neural network, which has 60 million parameters and 650,000 neurons, consists of five convolutional layers, some of which are followed by max-pooling layers, and three fully connected layers with a final 1000-way softmax. To make training faster, we used non-saturating neurons and a very efficient GPU implementation of the convolution operation. To reduce overfitting in the fully connected layers we employed a recently developed regularization method called "dropout" that proved to be very effective. We also entered a variant of this model in the ILSVRC-2012 competition and achieved a winning top-5 test error rate of 15.3%, compared to 26.2% achieved by the second-best entry.

12,532 Citations

Open accessJournal Article
Ronan Collobert, Jason Weston1, Léon Bottou, Michael Karlen  +2 moreInstitutions (3)
Abstract: We propose a unified neural network architecture and learning algorithm that can be applied to various natural language processing tasks including part-of-speech tagging, chunking, named entity recognition, and semantic role labeling. This versatility is achieved by trying to avoid task-specific engineering and therefore disregarding a lot of prior knowledge. Instead of exploiting man-made input features carefully optimized for each task, our system learns internal representations on the basis of vast amounts of mostly unlabeled training data. This work is then used as a basis for building a freely available tagging system with good performance and minimal computational requirements.

Topics: , ,

6,727 Citations

Open accessProceedings Article
01 Jan 2005-
Abstract: In this paper, we present bidirectional Long Short Term Memory (LSTM) networks, and a modified, full gradient version of the LSTM learning algorithm. We evaluate Bidirectional LSTM (BLSTM) and several other network architectures on the benchmark task of framewise phoneme classification, using the TIMIT database. Our main findings are that bidirectional networks outperform unidirectional ones, and Long Short Term Memory (LSTM) is much faster and also more accurate than both standard Recurrent Neural Nets (RNNs) and time-windowed Multilayer Perceptrons (MLPs). Our results support the view that contextual information is crucial to speech processing, and suggest that BLSTM is an effective architecture with which to exploit it'.

Topics:

3,028 Citations

Journal Article
08 Aug 2014-Science
Abstract: Inspired by the brain’s structure, we have developed an efficient, scalable, and flexible non–von Neumann architecture that leverages contemporary silicon technology. To demonstrate, we built a 5.4-billion-transistor chip with 4096 neurosynaptic cores interconnected via an intrachip network that integrates 1 million programmable spiking neurons and 256 million configurable synapses. Chips can be tiled in two dimensions via an interchip communication interface, seamlessly scaling the architecture to a cortexlike sheet of arbitrary size. The architecture is well suited to many applications that use complex neural networks in real time, for example, multiobject detection and classification. With 400-pixel-by-240-pixel video input at 30 frames per second, the chip consumes 63 milliwatts.