Advancing brain-inspired computing with hybrid neural networks.

Brain-inspired computing, drawing inspiration from the fundamental structure and information-processing mechanisms of the human brain, has gained significant momentum in recent years. It has emerged as a research paradigm centered on brain-computer dual-driven and multi-network integration. One noteworthy instance of this paradigm is the hybrid neural network (HNN), which integrates computer-science-oriented artificial neural networks (ANNs) with neuroscience-oriented spiking neural networks (SNNs). HNNs exhibit distinct advantages in various intelligent tasks, including perception, cognition and learning. This paper presents a comprehensive review of HNNs with an emphasis on their origin, concepts, biological perspective, construction framework and supporting systems. Furthermore, insights and suggestions for potential research directions are provided aiming to propel the advancement of the HNN paradigm.

Modularity Facilitates Classification Performance of Spiking Neural Networks for Decoding Cortical Spike Trains.

After the introduction of recurrence, an important property of the biological brain, spiking neural networks (SNNs) have achieved unprecedented classification performance. But they still cannot outperform many artificial neural networks. Modularity is another crucial feature of the biological brain. It remains unclear if modularity can also improve the performance of SNNs. To investigate this idea, we proposed the modular SNN, and compared its performance with a uniform SNN without modularity by employing them to classify cortical spike trains. For the first time, a significant improvement was found in our modular SNN. Further, we probed into the factors influencing the performance of the modular SNN and found: (a). The modular SNN outperformed the uniform SNN more significantly when the number of neurons in the networks increased; (b). The performance of the modular SNNs increased as the number of modules dropped. These preliminary but novel findings suggest that modularity may help develop better artificial intelligence and brain-machine interfaces. Also, the modular SNN may serve as a model for the study of neuronal spike synchrony.

Learnable axonal delay in spiking neural networks improves spoken word recognition.

Spiking neural networks (SNNs), which are composed of biologically plausible spiking neurons, and combined with bio-physically realistic auditory periphery models, offer a means to explore and understand human auditory processing-especially in tasks where precise timing is essential. However, because of the inherent temporal complexity in spike sequences, the performance of SNNs has remained less competitive compared to artificial neural networks (ANNs). To tackle this challenge, a fundamental research topic is the configuration of spike-timing and the exploration of more intricate architectures. In this work, we demonstrate a learnable axonal delay combined with local skip-connections yields state-of-the-art performance on challenging benchmarks for spoken word recognition. Additionally, we introduce an auxiliary loss term to further enhance accuracy and stability. Experiments on the neuromorphic speech benchmark datasets, NTIDIDIGITS and SHD, show improvements in performance when incorporating our delay module in comparison to vanilla feedforward SNNs. Specifically, with the integration of our delay module, the performance on NTIDIDIGITS and SHD improves by 14% and 18%, respectively. When paired with local skip-connections and the auxiliary loss, our approach surpasses both recurrent and convolutional neural networks, yet uses 10 × fewer parameters for NTIDIDIGITS and 7 × fewer for SHD.

Event-Driven Spiking Learning Algorithm Using Aggregated Labels.

Traditional spiking learning algorithm aims to train neurons to spike at a specific time or on a particular frequency, which requires precise time and frequency labels in the training process. While in reality, usually only aggregated labels of sequential patterns are provided. The aggregate-label (AL) learning is proposed to discover these predictive features in distracting background streams only by aggregated spikes. It has achieved much success recently, but it is still computationally intensive and has limited use in deep networks. To address these issues, we propose an event-driven spiking aggregate learning algorithm (SALA) in this article. Specifically, to reduce the computational complexity, we improve the conventional spike-threshold-surface (STS) calculation in AL learning by analytical calculating voltage peak values in spiking neurons. Then we derive the algorithm to multilayers by event-driven strategy using aggregated spikes. We conduct comprehensive experiments on various tasks including temporal clue recognition, segmented and continuous speech recognition, and neuromorphic image classification. The experimental results demonstrate that the new STS method improves the efficiency of AL learning significantly, and the proposed algorithm outperforms the conventional spiking algorithm in various temporal clue recognition tasks.

BIDL: a brain-inspired deep learning framework for spatiotemporal processing.

Brain-inspired deep spiking neural network (DSNN) which emulates the function of the biological brain provides an effective approach for event-stream spatiotemporal perception (STP), especially for dynamic vision sensor (DVS) signals. However, there is a lack of generalized learning frameworks that can handle various spatiotemporal modalities beyond event-stream, such as video clips and 3D imaging data. To provide a unified design flow for generalized spatiotemporal processing (STP) and to investigate the capability of lightweight STP processing via brain-inspired neural dynamics, this study introduces a training platform called brain-inspired deep learning (BIDL). This framework constructs deep neural networks, which leverage neural dynamics for processing temporal information and ensures high-accuracy spatial processing via artificial neural network layers. We conducted experiments involving various types of data, including video information processing, DVS information processing, 3D medical imaging classification, and natural language processing. These experiments demonstrate the efficiency of the proposed method. Moreover, as a research framework for researchers in the fields of neuroscience and machine learning, BIDL facilitates the exploration of different neural models and enables global-local co-learning. For easily fitting to neuromorphic chips and GPUs, the framework incorporates several optimizations, including iteration representation, state-aware computational graph, and built-in neural functions. This study presents a user-friendly and efficient DSNN builder for lightweight STP applications and has the potential to drive future advancements in bio-inspired research.

Named entity recognition for Chinese based on global pointer and adversarial training.

Named entity recognition aims to identify entities from unstructured text and is an important subtask for natural language processing and building knowledge graphs. Most of the existing entity recognition methods use conditional random fields as label decoders or use pointer networks for entity recognition. However, when the number of tags is large, the computational cost of method based on conditional random fields is high and the problem of nested entities cannot be solved. The pointer network uses two modules to identify the first and the last of the entities separately, and a single module can only focus on the information of the first or the last of the entities, but cannot pay attention to the global information of the entities. In addition, the neural network model has the problem of local instability. To solve mentioned problems, a named entity recognition model based on global pointer and adversarial training is proposed. To obtain global entity information, global pointer is used to decode entity information, and rotary relative position information is considered in the model designing to improve the model's perception of position; to solve the model's local instability problem, adversarial training is used to improve the robustness and generalization of the model. The experimental results show that the F1 score of the model are improved on several public datasets of OntoNotes5, MSRA, Resume, and Weibo compared with the existing mainstream models.

A Tandem Learning Rule for Effective Training and Rapid Inference of Deep Spiking Neural Networks.

Spiking neural networks (SNNs) represent the most prominent biologically inspired computing model for neuromorphic computing (NC) architectures. However, due to the nondifferentiable nature of spiking neuronal functions, the standard error backpropagation algorithm is not directly applicable to SNNs. In this work, we propose a tandem learning framework that consists of an SNN and an artificial neural network (ANN) coupled through weight sharing. The ANN is an auxiliary structure that facilitates the error backpropagation for the training of the SNN at the spike-train level. To this end, we consider the spike count as the discrete neural representation in the SNN and design an ANN neuronal activation function that can effectively approximate the spike count of the coupled SNN. The proposed tandem learning rule demonstrates competitive pattern recognition and regression capabilities on both the conventional frame- and event-based vision datasets, with at least an order of magnitude reduced inference time and total synaptic operations over other state-of-the-art SNN implementations. Therefore, the proposed tandem learning rule offers a novel solution to training efficient, low latency, and high-accuracy deep SNNs with low computing resources.

Polarized light compass decoding.

The brains of some insects can encode and decode polarization information and obtain heading angle information. Referring to the encoding ability of insects, exponential function encoding is designed to improve the stability of the polarized light compass artificial neural network. However, in the decoding process, only neurons with the largest activation degree are used for decoding (maximum value decoding), so the heading information contained in other neurons is not used. Therefore, average value decoding (AVD) and weighted AVD are proposed to use the heading information contained in multiple neurons to determine the heading. In addition, concerning the phenomenon of threshold activation of insect neurons, threshold value decoding (TVD) and weighted TVD are proposed, which can effectively eliminate the interference of neurons with low activation. Moreover, this paper proposes to improve the heading determination accuracy of the artificial neural network through pre-training. The simulation and experimental results show that the new, to the best of our knowledge, decoding methods and pre-training can effectively improve the heading determination accuracy of the artificial neural network.

Rectified Linear Postsynaptic Potential Function for Backpropagation in Deep Spiking Neural Networks.

Spiking neural networks (SNNs) use spatiotemporal spike patterns to represent and transmit information, which are not only biologically realistic but also suitable for ultralow-power event-driven neuromorphic implementation. Just like other deep learning techniques, deep SNNs (DeepSNNs) benefit from the deep architecture. However, the training of DeepSNNs is not straightforward because the well-studied error backpropagation (BP) algorithm is not directly applicable. In this article, we first establish an understanding as to why error BP does not work well in DeepSNNs. We then propose a simple yet efficient rectified linear postsynaptic potential function (ReL-PSP) for spiking neurons and a spike-timing-dependent BP (STDBP) learning algorithm for DeepSNNs where the timing of individual spikes is used to convey information (temporal coding), and learning (BP) is performed based on spike timing in an event-driven manner. We show that DeepSNNs trained with the proposed single spike time-based learning algorithm can achieve the state-of-the-art classification accuracy. Furthermore, by utilizing the trained model parameters obtained from the proposed STDBP learning algorithm, we demonstrate ultralow-power inference operations on a recently proposed neuromorphic inference accelerator. The experimental results also show that the neuromorphic hardware consumes 0.751 mW of the total power consumption and achieves a low latency of 47.71 ms to classify an image from the Modified National Institute of Standards and Technology (MNIST) dataset. Overall, this work investigates the contribution of spike timing dynamics for information encoding, synaptic plasticity, and decision-making, providing a new perspective to the design of future DeepSNNs and neuromorphic hardware.

Is Neuromorphic MNIST Neuromorphic? Analyzing the Discriminative Power of Neuromorphic Datasets in the Time Domain.

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?

HFNet: A CNN Architecture Co-designed for Neuromorphic Hardware With a Crossbar Array of Synapses.

The hardware-software co-optimization of neural network architectures is a field of research that emerged with the advent of commercial neuromorphic chips, such as the IBM TrueNorth and Intel Loihi. Development of simulation and automated mapping software tools in tandem with the design of neuromorphic hardware, whilst taking into consideration the hardware constraints, will play an increasingly significant role in deployment of system-level applications. This paper illustrates the importance and benefits of co-design of convolutional neural networks (CNN) that are to be mapped onto neuromorphic hardware with a crossbar array of synapses. Toward this end, we first study which convolution techniques are more hardware friendly and propose different mapping techniques for different convolutions. We show that, for a seven-layered CNN, our proposed mapping technique can reduce the number of cores used by 4.9-13.8 times for crossbar sizes ranging from 128 × 256 to 1,024 × 1,024, and this can be compared to the toeplitz method of mapping. We next develop an iterative co-design process for the systematic design of more hardware-friendly CNNs whilst considering hardware constraints, such as core sizes. A python wrapper, developed for the mapping process, is also useful for validating hardware design and studies on traffic volume and energy consumption. Finally, a new neural network dubbed HFNet is proposed using the above co-design process; it achieves a classification accuracy of 71.3% on the IMAGENET dataset (comparable to the VGG-16) but uses 11 times less cores for neuromorphic hardware with core size of 1,024 × 1,024. We also modified the HFNet to fit onto different core sizes and report on the corresponding classification accuracies. Various aspects of the paper are patent pending.

An Efficient and Perceptually Motivated Auditory Neural Encoding and Decoding Algorithm for Spiking Neural Networks.

The auditory front-end is an integral part of a spiking neural network (SNN) when performing auditory cognitive tasks. It encodes the temporal dynamic stimulus, such as speech and audio, into an efficient, effective and reconstructable spike pattern to facilitate the subsequent processing. However, most of the auditory front-ends in current studies have not made use of recent findings in psychoacoustics and physiology concerning human listening. In this paper, we propose a neural encoding and decoding scheme that is optimized for audio processing. The neural encoding scheme, that we call Biologically plausible Auditory Encoding (BAE), emulates the functions of the perceptual components of the human auditory system, that include the cochlear filter bank, the inner hair cells, auditory masking effects from psychoacoustic models, and the spike neural encoding by the auditory nerve. We evaluate the perceptual quality of the BAE scheme using PESQ; the performance of the BAE based on sound classification and speech recognition experiments. Finally, we also built and published two spike-version of speech datasets: the Spike-TIDIGITS and the Spike-TIMIT, for researchers to use and benchmarking of future SNN research.

A Spiking Neural Network Framework for Robust Sound Classification.

Environmental sounds form part of our daily life. With the advancement of deep learning models and the abundance of training data, the performance of automatic sound classification (ASC) systems has improved significantly in recent years. However, the high computational cost, hence high power consumption, remains a major hurdle for large-scale implementation of ASC systems on mobile and wearable devices. Motivated by the observations that humans are highly effective and consume little power whilst analyzing complex audio scenes, we propose a biologically plausible ASC framework, namely SOM-SNN. This framework uses the unsupervised self-organizing map (SOM) for representing frequency contents embedded within the acoustic signals, followed by an event-based spiking neural network (SNN) for spatiotemporal spiking pattern classification. We report experimental results on the RWCP environmental sound and TIDIGITS spoken digits datasets, which demonstrate competitive classification accuracies over other deep learning and SNN-based models. The SOM-SNN framework is also shown to be highly robust to corrupting noise after multi-condition training, whereby the model is trained with noise-corrupted sound samples. Moreover, we discover the early decision making capability of the proposed framework: an accurate classification can be made with an only partial presentation of the input.

Effects of Calcium Spikes in the Layer 5 Pyramidal Neuron on Coincidence Detection and Activity Propagation.

The role of dendritic spiking mechanisms in neural processing is so far poorly understood. To investigate the role of calcium spikes in the functional properties of the single neuron and recurrent networks, we investigated a three compartment neuron model of the layer 5 pyramidal neuron with calcium dynamics in the distal compartment. By performing single neuron simulations with noisy synaptic input and occasional large coincident input at either just the distal compartment or at both somatic and distal compartments, we show that the presence of calcium spikes confers a substantial advantage for coincidence detection in the former case and a lesser advantage in the latter. We further show that the experimentally observed critical frequency phenomenon, in which action potentials triggered by stimuli near the soma above a certain frequency trigger a calcium spike at distal dendrites, leading to further somatic depolarization, is not exhibited by a neuron receiving realistically noisy synaptic input, and so is unlikely to be a necessary component of coincidence detection. We next investigate the effect of calcium spikes in propagation of spiking activities in a feed-forward network (FFN) embedded in a balanced recurrent network. The excitatory neurons in the network are again connected to either just the distal, or both somatic and distal compartments. With purely distal connectivity, activity propagation is stable and distinguishable for a large range of recurrent synaptic strengths if the feed-forward connections are sufficiently strong, but propagation does not occur in the absence of calcium spikes. When connections are made to both the somatic and the distal compartments, activity propagation is achieved for neurons with active calcium dynamics at a much smaller number of neurons per pool, compared to a network of passive neurons, but quickly becomes unstable as the strength of recurrent synapses increases. Activity propagation at higher scaling factors can be stabilized by increasing network inhibition or introducing short term depression in the excitatory synapses, but the signal to noise ratio remains low. Our results demonstrate that the interaction of synchrony with dendritic spiking mechanisms can have profound consequences for the dynamics on the single neuron and network level.

Modeling the calcium spike as a threshold triggered fixed waveform for synchronous inputs in the fluctuation regime.

Modeling the layer 5 pyramidal neuron as a system of three connected isopotential compartments, the soma, proximal, and distal compartment, with calcium spike dynamics in the distal compartment following first order kinetics, we are able to reproduce in-vitro experimental results which demonstrate the involvement of calcium spikes in action potentials generation. To explore how calcium spikes affect the neuronal output in-vivo, we emulate in-vivo like conditions by embedding the neuron model in a regime of low background fluctuations with occasional large synchronous inputs. In such a regime, a full calcium spike is only triggered by the synchronous events in a threshold like manner and has a stereotypical waveform. Hence, in such a regime, we are able to replace the calcium dynamics with a simpler threshold triggered current of fixed waveform, which is amenable to analytical treatment. We obtain analytically the mean somatic membrane potential excursion due to a calcium spike being triggered while in the fluctuating regime. Our analytical form that accounts for the covariance between conductances and the membrane potential shows a better agreement with simulation results than a naive first order approximation.