Spike-based strategies for rapid processing

Focal seizures induce spatiotemporally organized spiking activity in the human cortex

Epileptic seizures are debilitating because of the clinical symptoms they produce. These symptoms, in turn, may stem directly from disruptions in neural coding. Recent evidence has suggested that the specific temporal order, or sequence, of spiking across a population of cortical neurons may encode information. Here, we investigate how seizures disrupt neuronal spiking sequences in the human brain by recording multi-unit activity from the cerebral cortex in five male participants undergoing monitoring for seizures. We find that pathological discharges during seizures are associated with bursts of spiking activity across a population of cortical neurons. These bursts are organized into highly consistent and stereotyped temporal sequences. As the seizure evolves, spiking sequences diverge from the sequences observed at baseline and become more spatially organized. The direction of this spatial organization matches the direction of the ictal discharges, which spread over the cortex as traveling waves. Our data therefore suggest that seizures can entrain cortical spiking sequences by changing the spatial organization of neuronal firing, providing a possible mechanism by which seizures create symptoms.

High-performance deep spiking neural networks with 0.3 spikes per neuron

Communication by rare, binary spikes is a key factor for the energy efficiency of biological brains. However, it is harder to train biologically-inspired spiking neural networks than artificial neural networks. This is puzzling given that theoretical results provide exact mapping algorithms from artificial to spiking neural networks with time-to-first-spike coding. In this paper we analyze in theory and simulation the learning dynamics of time-to-first-spike-networks and identify a specific instance of the vanishing-or-exploding gradient problem. While two choices of spiking neural network mappings solve this problem at initialization, only the one with a constant slope of the neuron membrane potential at threshold guarantees the equivalence of the training trajectory between spiking and artificial neural networks with rectified linear units. For specific image classification architectures comprising feed-forward dense or convolutional layers, we demonstrate that deep spiking neural network models can be effectively trained from scratch on MNIST and Fashion-MNIST datasets, or fine-tuned on large-scale datasets, such as CIFAR10, CIFAR100 and PLACES365, to achieve the exact same performance as that of artificial neural networks, surpassing previous spiking neural networks. Our approach accomplishes high-performance classification with less than 0.3 spikes per neuron, lending itself for an energy-efficient implementation. We also show that fine-tuning spiking neural networks with our robust gradient descent algorithm enables their optimization for hardware implementations with low latency and resilience to noise and quantization.

Paired competing neurons improving STDP supervised local learning in Spiking Neural Networks

Direct training of Spiking Neural Networks (SNNs) on neuromorphic hardware has the potential to significantly reduce the energy consumption of artificial neural network training. SNNs trained with Spike Timing-Dependent Plasticity (STDP) benefit from gradient-free and unsupervised local learning, which can be easily implemented on ultra-low-power neuromorphic hardware. However, classification tasks cannot be performed solely with unsupervised STDP. In this paper, we propose Stabilized Supervised STDP (S2-STDP), a supervised STDP learning rule to train the classification layer of an SNN equipped with unsupervised STDP for feature extraction. S2-STDP integrates error-modulated weight updates that align neuron spikes with desired timestamps derived from the average firing time within the layer. Then, we introduce a training architecture called Paired Competing Neurons (PCN) to further enhance the learning capabilities of our classification layer trained with S2-STDP. PCN associates each class with paired neurons and encourages neuron specialization toward target or non-target samples through intra-class competition. We evaluate our methods on image recognition datasets, including MNIST, Fashion-MNIST, and CIFAR-10. Results show that our methods outperform state-of-the-art supervised STDP learning rules, for comparable architectures and numbers of neurons. Further analysis demonstrates that the use of PCN enhances the performance of S2-STDP, regardless of the hyperparameter set and without introducing any additional hyperparameters.

The mechanism of human color vision and potential implanted devices for artificial color vision

Vision plays a major role in perceiving external stimuli and information in our daily lives. The neural mechanism of color vision is complicated, involving the co-ordinated functions of a variety of cells, such as retinal cells and lateral geniculate nucleus cells, as well as multiple levels of the visual cortex. In this work, we reviewed the history of experimental and theoretical studies on this issue, from the fundamental functions of the individual cells of the visual system to the coding in the transmission of neural signals and sophisticated brain processes at different levels. We discuss various hypotheses, models, and theories related to the color vision mechanism and present some suggestions for developing novel implanted devices that may help restore color vision in visually impaired people or introduce artificial color vision to those who need it.

Characterizing the aggregated encoding method utilizing bursts activated by a VCSEL-neuron with a feedback structure

The rapid advancement of photonic technologies has facilitated the development of photonic neurons that emulate neuronal functionalities akin to those observed in the human brain. Neuronal bursts frequently occur in behaviors where information is encoded and transmitted. Here, we present the demonstration of the bursting response activated by an artificial photonic neuron. This neuron utilizes a single vertical-cavity surface-emitting laser (VCSEL) and encodes multiple stimuli effectively by varying the spike count during a burst based on the polarization competition in the VCSEL. By virtue of the modulated optical injection in the VCSEL employed to trigger the spiking response, we activate bursts output in the VCSEL with a feedback structure in this scheme. The bursting response activated by the VCSEL-neuron exhibits neural signal characteristics, promising an excitation threshold and the refractory period. Significantly, this marks the inaugural implementation of a controllable integrated encoding scheme predicated on bursts within photonic neurons. There are two remarkable merits; on the one hand, the interspike interval of bursts is distinctly diminished, amounting to merely one twenty-fourth compared to that observed in optoelectronic oscillators. Moreover, the interspike period of bursts is about 70.8% shorter than the period of spikes activated by a VCSEL neuron without optical feedback. Our results may shed light on the analogy between optical and biological neurons and open the door to fast burst encoding-based optical systems with a speed several orders of magnitude faster than their biological counterparts.

A Hybrid Neural Coding Approach for Pattern Recognition With Spiking Neural Networks

Recently, brain-inspired spiking neural networks (SNNs) have demonstrated promising capabilities in solving pattern recognition tasks. However, these SNNs are grounded on homogeneous neurons that utilize a uniform neural coding for information representation. Given that each neural coding scheme possesses its own merits and drawbacks, these SNNs encounter challenges in achieving optimal performance such as accuracy, response time, efficiency, and robustness, all of which are crucial for practical applications. In this study, we argue that SNN architectures should be holistically designed to incorporate heterogeneous coding schemes. As an initial exploration in this direction, we propose a hybrid neural coding and learning framework, which encompasses a neural coding zoo with diverse neural coding schemes discovered in neuroscience. Additionally, it incorporates a flexible neural coding assignment strategy to accommodate task-specific requirements, along with novel layer-wise learning methods to effectively implement hybrid coding SNNs. We demonstrate the superiority of the proposed framework on image classification and sound localization tasks. Specifically, the proposed hybrid coding SNNs achieve comparable accuracy to state-of-the-art SNNs, while exhibiting significantly reduced inference latency and energy consumption, as well as high noise robustness. This study yields valuable insights into hybrid neural coding designs, paving the way for developing high-performance neuromorphic systems.

Enhanced Release Probability without Changes in Synaptic Delay during Analogue-Digital Facilitation

Neuronal timing with millisecond precision is critical for many brain functions such as sensory perception, learning and memory formation. At the level of the chemical synapse, the synaptic delay is determined by the presynaptic release probability (Pr) and the waveform of the presynaptic action potential (AP). For instance, paired-pulse facilitation or presynaptic long-term potentiation are associated with reductions in the synaptic delay, whereas paired-pulse depression or presynaptic long-term depression are associated with an increased synaptic delay. Parallelly, the AP broadening that results from the inactivation of voltage gated potassium (Kv) channels responsible for the repolarization phase of the AP delays the synaptic response, and the inactivation of sodium (Nav) channels by voltage reduces the synaptic latency. However, whether synaptic delay is modulated during depolarization-induced analogue-digital facilitation (d-ADF), a form of context-dependent synaptic facilitation induced by prolonged depolarization of the presynaptic neuron and mediated by the voltage-inactivation of presynaptic Kv1 channels, remains unclear. We show here that despite Pr being elevated during d-ADF at pyramidal L5-L5 cell synapses, the synaptic delay is surprisingly unchanged. This finding suggests that both Pr- and AP-dependent changes in synaptic delay compensate for each other during d-ADF. We conclude that, in contrast to other short- or long-term modulations of presynaptic release, synaptic timing is not affected during d-ADF because of the opposite interaction of Pr- and AP-dependent modulations of synaptic delay.

A fully spiking coupled model of a deep neural network and a recurrent attractor explains dynamics of decision making in an object recognition task

Objective.Object recognition and making a choice regarding the recognized object is pivotal for most animals. This process in the brain contains information representation and decision making steps which both take different amount of times for different objects. While dynamics of object recognition and decision making are usually ignored in object recognition models, here we proposed a fully spiking hierarchical model, explaining the process of object recognition from information representation to making decision.Approach.Coupling a deep neural network and a recurrent attractor based decision making model beside using spike time dependent plasticity learning rules in several convolutional and pooling layers, we proposed a model which can resemble brain behaviors during an object recognition task. We also measured human choices and reaction times in a psychophysical object recognition task and used it as a reference to evaluate the model.Main results.The proposed model explains not only the probability of making a correct decision but also the time that it takes to make a decision. Importantly, neural firing rates in both feature representation and decision making levels mimic the observed patterns in animal studies (number of spikes (p-value < 10-173) and the time of the peak response (p-value < 10-31) are significantly modulated with the strength of the stimulus). Moreover, the speed-accuracy trade-off as a well-known characteristic of decision making process in the brain is also observed in the model (changing the decision bound significantly affect the reaction time (p-value < 10-59) and accuracy (p-value < 10-165)).Significance.We proposed a fully spiking deep neural network which can explain dynamics of making decision about an object in both neural and behavioral level. Results showed that there is a strong and significant correlation (r= 0.57) between the reaction time of the model and of human participants in the psychophysical object recognition task.

Exact solutions for signal propagation along an excitable transmission line

A simple transmission line composed of pulse-coupled units is presented. The model captures the basic properties of excitable media with, in particular, the robust transmission of information via traveling wave solutions. For rectified linear units with a cut-off threshold, the model is exactly solvable and analytical results on propagation are presented. The ability to convey a nontrivial message is studied in detail.

Sparse-firing regularization methods for spiking neural networks with time-to-first-spike coding

The training of multilayer spiking neural networks (SNNs) using the error backpropagation algorithm has made significant progress in recent years. Among the various training schemes, the error backpropagation method that directly uses the firing time of neurons has attracted considerable attention because it can realize ideal temporal coding. This method uses time-to-first-spike (TTFS) coding, in which each neuron fires at most once, and this restriction on the number of firings enables information to be processed at a very low firing frequency. This low firing frequency increases the energy efficiency of information processing in SNNs. However, only an upper limit has been provided for TTFS-coded SNNs, and the information-processing capability of SNNs at lower firing frequencies has not been fully investigated. In this paper, we propose two spike-timing-based sparse-firing (SSR) regularization methods to further reduce the firing frequency of TTFS-coded SNNs. Both methods are characterized by the fact that they only require information about the firing timing and associated weights. The effects of these regularization methods were investigated on the MNIST, Fashion-MNIST, and CIFAR-10 datasets using multilayer perceptron networks and convolutional neural network structures.

The Influence of Phosphene Synchrony in Driving Object Binding in a Simulation of Artificial Vision

Purpose

Electrical microstimulation techniques used in visual prostheses are designed to restore visual function following acquired blindness. Patterns of induced focal percepts, known as phosphenes, are achieved by applying localized electrical pulses to the visual pathway to bypass the impaired site in order to convey images from the external world. Here, we use a simulation of artificial vision to manipulate relationships between individual phosphenes to observe the effects on object binding and perception. We hypothesize that synchronous phosphene presentation will facilitate object perception as compared to asynchronous presentation.

Methods

A model system that tracks gaze position of normal, sighted participants to present patterns of phosphenes on a computer screen was used to simulate prosthetic vision. Participants performed a reading task at varying font sizes (1.1-1.4 logMAR) and under varying levels of phosphene temporal noise while reading accuracy and speed were measured.

Results

Reading performance was significantly affected by temporal noise in phosphene presentation, with increasing desynchronization leading to lower reading scores. A drop in performance was also observed when the total latency between the gaze position and phosphene update was increased without adding temporal noise.

Conclusions

Object perception (here, text perception) is enhanced with synchronously presented phosphenes as compared to asynchronously presented ones. These results are fundamental for developing an efficient temporal pattern of stimulation and for the creation of high-fidelity prosthetic vision.

An exact mapping from ReLU networks to spiking neural networks

Deep spiking neural networks (SNNs) offer the promise of low-power artificial intelligence. However, training deep SNNs from scratch or converting deep artificial neural networks to SNNs without loss of performance has been a challenge. Here we propose an exact mapping from a network with Rectified Linear Units (ReLUs) to an SNN that fires exactly one spike per neuron. For our constructive proof, we assume that an arbitrary multi-layer ReLU network with or without convolutional layers, batch normalization and max pooling layers was trained to high performance on some training set. Furthermore, we assume that we have access to a representative example of input data used during training and to the exact parameters (weights and biases) of the trained ReLU network. The mapping from deep ReLU networks to SNNs causes zero percent drop in accuracy on CIFAR10, CIFAR100 and the ImageNet-like data sets Places365 and PASS. More generally our work shows that an arbitrary deep ReLU network can be replaced by an energy-efficient single-spike neural network without any loss of performance.

Adaptive STDP-based on-chip spike pattern detection

A spiking neural network (SNN) is a bottom-up tool used to describe information processing in brain microcircuits. It is becoming a crucial neuromorphic computational model. Spike-timing-dependent plasticity (STDP) is an unsupervised brain-like learning rule implemented in many SNNs and neuromorphic chips. However, a significant performance gap exists between ideal model simulation and neuromorphic implementation. The performance of STDP learning in neuromorphic chips deteriorates because the resolution of synaptic efficacy in such chips is generally restricted to 6 bits or less, whereas simulations employ the entire 64-bit floating-point precision available on digital computers. Previously, we introduced a bio-inspired learning rule named adaptive STDP and demonstrated via numerical simulation that adaptive STDP (using only 4-bit fixed-point synaptic efficacy) performs similarly to STDP learning (using 64-bit floating-point precision) in a noisy spike pattern detection model. Herein, we present the experimental results demonstrating the performance of adaptive STDP learning. To the best of our knowledge, this is the first study that demonstrates unsupervised noisy spatiotemporal spike pattern detection to perform well and maintain the simulation performance on a mixed-signal CMOS neuromorphic chip with low-resolution synaptic efficacy. The chip was designed in Taiwan Semiconductor Manufacturing Company (TSMC) 250 nm CMOS technology node and comprises a soma circuit and 256 synapse circuits along with their learning circuitry.

Online spike-based recognition of digits with ultrafast microlaser neurons

Classification and recognition tasks performed on photonic hardware-based neural networks often require at least one offline computational step, such as in the increasingly popular reservoir computing paradigm. Removing this offline step can significantly improve the response time and energy efficiency of such systems. We present numerical simulations of different algorithms that utilize ultrafast photonic spiking neurons as receptive fields to allow for image recognition without an offline computing step. In particular, we discuss the merits of event, spike-time and rank-order based algorithms adapted to this system. These techniques have the potential to significantly improve the efficiency and effectiveness of optical classification systems, minimizing the number of spiking nodes required for a given task and leveraging the parallelism offered by photonic hardware.

SpikeShip: A method for fast, unsupervised discovery of high-dimensional neural spiking patterns

Neural coding and memory formation depend on temporal spiking sequences that span high-dimensional neural ensembles. The unsupervised discovery and characterization of these spiking sequences requires a suitable dissimilarity measure to spiking patterns, which can then be used for clustering and decoding. Here, we present a new dissimilarity measure based on optimal transport theory called SpikeShip, which compares multi-neuron spiking patterns based on all the relative spike-timing relationships among neurons. SpikeShip computes the optimal transport cost to make all the relative spike-timing relationships (across neurons) identical between two spiking patterns. We show that this transport cost can be decomposed into a temporal rigid translation term, which captures global latency shifts, and a vector of neuron-specific transport flows, which reflect inter-neuronal spike timing differences. SpikeShip can be effectively computed for high-dimensional neuronal ensembles, has a low (linear) computational cost that has the same order as the spike count, and is sensitive to higher-order correlations. Furthermore, SpikeShip is binless, can handle any form of spike time distributions, is not affected by firing rate fluctuations, can detect patterns with a low signal-to-noise ratio, and can be effectively combined with a sliding window approach. We compare the advantages and differences between SpikeShip and other measures like SPIKE and Victor-Purpura distance. We applied SpikeShip to large-scale Neuropixel recordings during spontaneous activity and visual encoding. We show that high-dimensional spiking sequences detected via SpikeShip reliably distinguish between different natural images and different behavioral states. These spiking sequences carried complementary information to conventional firing rate codes. SpikeShip opens new avenues for studying neural coding and memory consolidation by rapid and unsupervised detection of temporal spiking patterns in high-dimensional neural ensembles.

On-Chip Adaptive Implementation of Neuromorphic Spiking Sensory Systems with Self-X Capabilities

In contemporary devices, the number and diversity of sensors is increasing, thus, requiring both efficient and robust interfacing to the sensors. Implementing the interfacing systems in advanced integration technologies faces numerous issues due to manufacturing deviations, signal swings, noise, etc. The interface sensor designers escape to the time domain and digital design techniques to handle these challenges. Biology gives examples of efficient machines that have vastly outperformed conventional technology. This work pursues a neuromorphic spiking sensory system design with the same efficient style as biology. Our chip, that comprises the essential elements of the adaptive neuromorphic spiking sensory system, such as the neuron, synapse, adaptive coincidence detection (ACD), and self-adaptive spike-to-rank coding (SA-SRC), was manufactured in XFAB CMOS 0.35 μm technology via EUROPRACTICE. The main emphasis of this paper is to present the measurement outcomes of the SA-SRC on-chip, evaluating the efficacy of its adaptation scheme, and assessing its capability to produce spike orders that correspond to the temporal difference between the two spikes received at its inputs. The SA-SRC plays a crucial role in performing the primary function of the adaptive neuromorphic spiking sensory system. The measurement results of the chip confirm the simulation results of our previous work.

Robust encoding of natural stimuli by neuronal response sequences in monkey visual cortex

Parallel multisite recordings in the visual cortex of trained monkeys revealed that the responses of spatially distributed neurons to natural scenes are ordered in sequences. The rank order of these sequences is stimulus-specific and maintained even if the absolute timing of the responses is modified by manipulating stimulus parameters. The stimulus specificity of these sequences was highest when they were evoked by natural stimuli and deteriorated for stimulus versions in which certain statistical regularities were removed. This suggests that the response sequences result from a matching operation between sensory evidence and priors stored in the cortical network. Decoders trained on sequence order performed as well as decoders trained on rate vectors but the former could decode stimulus identity from considerably shorter response intervals than the latter. A simulated recurrent network reproduced similarly structured stimulus-specific response sequences, particularly once it was familiarized with the stimuli through non-supervised Hebbian learning. We propose that recurrent processing transforms signals from stationary visual scenes into sequential responses whose rank order is the result of a Bayesian matching operation. If this temporal code were used by the visual system it would allow for ultrafast processing of visual scenes.

Spontaneous activity in cortical neurons is stereotyped and non-Poisson

Neurons fire even in the absence of sensory stimulation or task demands. Numerous theoretical studies have modeled this spontaneous activity as a Poisson process with uncorrelated intervals between successive spikes and a variance in firing rate equal to the mean. Experimental tests of this hypothesis have yielded variable results, though most have concluded that firing is not Poisson. However, these tests say little about the ways firing might deviate from randomness. Nor are they definitive because many different distributions can have equal means and variances. Here, we characterized spontaneous spiking patterns in extracellular recordings from monkey, cat, and mouse cerebral cortex neurons using rate-normalized spike train autocorrelation functions (ACFs) and a logarithmic timescale. If activity was Poisson, this function should be flat. This was almost never the case. Instead, ACFs had diverse shapes, often with characteristic peaks in the 1-700 ms range. Shapes were stable over time, up to the longest recording periods used (51 min). They did not fall into obvious clusters. ACFs were often unaffected by visual stimulation, though some abruptly changed during brain state shifts. These behaviors may have their origin in the intrinsic biophysics and dendritic anatomy of the cells or in the inputs they receive.

VTSNN: a virtual temporal spiking neural network

Spiking neural networks (SNNs) have recently demonstrated outstanding performance in a variety of high-level tasks, such as image classification. However, advancements in the field of low-level assignments, such as image reconstruction, are rare. This may be due to the lack of promising image encoding techniques and corresponding neuromorphic devices designed specifically for SNN-based low-level vision problems. This paper begins by proposing a simple yet effective undistorted weighted-encoding-decoding technique, which primarily consists of an Undistorted Weighted-Encoding (UWE) and an Undistorted Weighted-Decoding (UWD). The former aims to convert a gray image into spike sequences for effective SNN learning, while the latter converts spike sequences back into images. Then, we design a new SNN training strategy, known as Independent-Temporal Backpropagation (ITBP) to avoid complex loss propagation in spatial and temporal dimensions, and experiments show that ITBP is superior to Spatio-Temporal Backpropagation (STBP). Finally, a so-called Virtual Temporal SNN (VTSNN) is formulated by incorporating the above-mentioned approaches into U-net network architecture, fully utilizing the potent multiscale representation capability. Experimental results on several commonly used datasets such as MNIST, F-MNIST, and CIFAR10 demonstrate that the proposed method produces competitive noise-removal performance extremely which is superior to the existing work. Compared to ANN with the same architecture, VTSNN has a greater chance of achieving superiority while consuming ~1/274 of the energy. Specifically, using the given encoding-decoding strategy, a simple neuromorphic circuit could be easily constructed to maximize this low-carbon strategy.

Doppler Frequency-Shift Information Processing in WOx -Based Memristive Synapse for Auditory Motion Perception

Auditory motion perception is one crucial capability to decode and discriminate the spatiotemporal information for neuromorphic auditory systems. Doppler frequency-shift feature and interaural time difference (ITD) are two fundamental cues of auditory information processing. In this work, the functions of azimuth detection and velocity detection, as the typical auditory motion perception, are demonstrated in a WO_x -based memristive synapse. The WO_x memristor presents both the volatile mode (M1) and semi-nonvolatile mode (M2), which are capable of implementing the high-pass filtering and processing the spike trains with a relative timing and frequency shift. In particular, the Doppler frequency-shift information processing for velocity detection is emulated in the WO_x memristor based auditory system for the first time, which relies on a scheme of triplet spike-timing-dependent-plasticity in the memristor. These results provide new opportunities for the mimicry of auditory motion perception and enable the auditory sensory system to be applied in future neuromorphic sensing.

Resolving the neural mechanism of core object recognition in space and time: A computational approach

The underlying mechanism of object recognition- a fundamental brain ability- has been investigated in various studies. However, balancing between the speed and accuracy of recognition is less explored. Most of the computational models of object recognition are not potentially able to explain the recognition time and, thus, only focus on the recognition accuracy because of two reasons: lack of a temporal representation mechanism for sensory processing and using non-biological classifiers for decision-making processing. Here, we proposed a hierarchical temporal model of object recognition using a spiking deep neural network coupled to a biologically plausible decision-making model for explaining both recognition time and accuracy. We showed that the response dynamics of the proposed model can resemble those of the brain. Firstly, in an object recognition task, the model can mimic human's and monkey's recognition time as well as accuracy. Secondly, the model can replicate different speed-accuracy trade-off regimes as observed in the literature. More importantly, we demonstrated that temporal representation of different abstraction levels (superordinate, midlevel, and subordinate) in the proposed model matched the brain representation dynamics observed in previous studies. We conclude that the accumulation of spikes, generated by a hierarchical feedforward spiking structure, to reach abound can well explain not even the dynamics of making a decision, but also the representations dynamics for different abstraction levels.

Saccades are locked to the phase of alpha oscillations during natural reading

We saccade 3 to 5 times per second when reading. However, little is known about the neuronal mechanisms coordinating the oculomotor and visual systems during such rapid processing. Here, we ask if brain oscillations play a role in the temporal coordination of the visuomotor integration. We simultaneously acquired MEG and eye-tracking data while participants read sentences silently. Every sentence was embedded with a target word of either high or low lexical frequency. Our key finding demonstrated that saccade onsets were locked to the phase of alpha oscillations (8 to 13 Hz), and in particular, for saccades towards low frequency words. Source modelling demonstrated that the alpha oscillations to which the saccades were locked, were generated in the right-visual motor cortex (BA 7). Our findings suggest that the alpha oscillations serve to time the processing between the oculomotor and visual systems during natural reading, and that this coordination becomes more pronounced for demanding words.

Integration of neuromorphic AI in event-driven distributed digitized systems: Concepts and research directions

Increasing complexity and data-generation rates in cyber-physical systems and the industrial Internet of things are calling for a corresponding increase in AI capabilities at the resource-constrained edges of the Internet. Meanwhile, the resource requirements of digital computing and deep learning are growing exponentially, in an unsustainable manner. One possible way to bridge this gap is the adoption of resource-efficient brain-inspired "neuromorphic" processing and sensing devices, which use event-driven, asynchronous, dynamic neurosynaptic elements with colocated memory for distributed processing and machine learning. However, since neuromorphic systems are fundamentally different from conventional von Neumann computers and clock-driven sensor systems, several challenges are posed to large-scale adoption and integration of neuromorphic devices into the existing distributed digital-computational infrastructure. Here, we describe the current landscape of neuromorphic computing, focusing on characteristics that pose integration challenges. Based on this analysis, we propose a microservice-based conceptual framework for neuromorphic systems integration, consisting of a neuromorphic-system proxy, which would provide virtualization and communication capabilities required in distributed systems of systems, in combination with a declarative programming approach offering engineering-process abstraction. We also present concepts that could serve as a basis for the realization of this framework, and identify directions for further research required to enable large-scale system integration of neuromorphic devices.

Direct Training via Backpropagation for Ultra-Low-Latency Spiking Neural Networks with Multi-Threshold

Spiking neural networks (SNNs) can utilize spatio-temporal information and have the characteristic of energy efficiency, being a good alternative to deep neural networks (DNNs). The event-driven information processing means that SNNs can reduce the expensive computation of DNNs and save a great deal of energy consumption. However, high training and inference latency is a limitation of the development of deeper SNNs. SNNs usually need tens or even hundreds of time steps during the training and inference process, which causes not only an increase in latency but also excessive energy consumption. To overcome this problem, we propose a novel training method based on backpropagation (BP) for ultra-low-latency (1–2 time steps) SNNs with multi-threshold. In order to increase the information capacity of each spike, we introduce the multi-threshold Leaky Integrate and Fired (LIF) model. The experimental results show that our proposed method achieves average accuracy of 99.56%, 93.08%, and 87.90% on MNIST, FashionMNIST, and CIFAR10, respectively, with only two time steps. For the CIFAR10 dataset, our proposed method achieves 1.12% accuracy improvement over the previously reported directly trained SNNs with fewer time steps.

Neuromorphic Tactile Edge Orientation Classification in an Unsupervised Spiking Neural Network

Dexterous manipulation in robotic hands relies on an accurate sense of artificial touch. Here we investigate neuromorphic tactile sensation with an event-based optical tactile sensor combined with spiking neural networks for edge orientation detection. The sensor incorporates an event-based vision system (mini-eDVS) into a low-form factor artificial fingertip (the NeuroTac). The processing of tactile information is performed through a Spiking Neural Network with unsupervised Spike-Timing-Dependent Plasticity (STDP) learning, and the resultant output is classified with a 3-nearest neighbours classifier. Edge orientations were classified in 10-degree increments while tapping vertically downward and sliding horizontally across the edge. In both cases, we demonstrate that the sensor is able to reliably detect edge orientation, and could lead to accurate, bio-inspired, tactile processing in robotics and prosthetics applications.

Digital computing through randomness and order in neural networks

We propose that coding and decoding in the brain are achieved through digital computation using three principles: relative ordinal coding of inputs, random connections between neurons, and belief voting. Due to randomization and despite the coarseness of the relative codes, we show that these principles are sufficient for coding and decoding sequences with error-free reconstruction. In particular, the number of neurons needed grows linearly with the size of the input repertoire growing exponentially. We illustrate our model by reconstructing sequences with repertoires on the order of a billion items. From this, we derive the Shannon equations for the capacity limit to learn and transfer information in the neural population, which is then generalized to any type of neural network. Following the maximum entropy principle of efficient coding, we show that random connections serve to decorrelate redundant information in incoming signals, creating more compact codes for neurons and therefore, conveying a larger amount of information. Henceforth, despite the unreliability of the relative codes, few neurons become necessary to discriminate the original signal without error. Finally, we discuss the significance of this digital computation model regarding neurobiological findings in the brain and more generally with artificial intelligence algorithms, with a view toward a neural information theory and the design of digital neural networks.

Time Is of the Essence: Neural Codes, Synchronies, Oscillations, Architectures

Time is of the essence in how neural codes, synchronies, and oscillations might function in encoding, representation, transmission, integration, storage, and retrieval of information in brains. This Hypothesis and Theory article examines observed and possible relations between codes, synchronies, oscillations, and types of neural networks they require. Toward reverse-engineering informational functions in brains, prospective, alternative neural architectures incorporating principles from radio modulation and demodulation, active reverberant circuits, distributed content-addressable memory, signal-signal time-domain correlation and convolution operations, spike-correlation-based holography, and self-organizing, autoencoding anticipatory systems are outlined. Synchronies and oscillations are thought to subserve many possible functions: sensation, perception, action, cognition, motivation, affect, memory, attention, anticipation, and imagination. These include direct involvement in coding attributes of events and objects through phase-locking as well as characteristic patterns of spike latency and oscillatory response. They are thought to be involved in segmentation and binding, working memory, attention, gating and routing of signals, temporal reset mechanisms, inter-regional coordination, time discretization, time-warping transformations, and support for temporal wave-interference based operations. A high level, partial taxonomy of neural codes consists of channel, temporal pattern, and spike latency codes. The functional roles of synchronies and oscillations in candidate neural codes, including oscillatory phase-offset codes, are outlined. Various forms of multiplexing neural signals are considered: time-division, frequency-division, code-division, oscillatory-phase, synchronized channels, oscillatory hierarchies, polychronous ensembles. An expandable, annotative neural spike train framework for encoding low- and high-level attributes of events and objects is proposed. Coding schemes require appropriate neural architectures for their interpretation. Time-delay, oscillatory, wave-interference, synfire chain, polychronous, and neural timing networks are discussed. Some novel concepts for formulating an alternative, more time-centric theory of brain function are discussed. As in radio communication systems, brains can be regarded as networks of dynamic, adaptive transceivers that broadcast and selectively receive multiplexed temporally-patterned pulse signals. These signals enable complex signal interactions that select, reinforce, and bind common subpatterns and create emergent lower dimensional signals that propagate through spreading activation interference networks. If memory traces share the same kind of temporal pattern forms as do active neuronal representations, then distributed, holograph-like content-addressable memories are made possible via temporal pattern resonances.

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.

Advancements in Algorithms and Neuromorphic Hardware for Spiking Neural Networks

Artificial neural networks (ANNs) have experienced a rapid advancement for their success in various application domains, including autonomous driving and drone vision. Researchers have been improving the performance efficiency and computational requirement of ANNs inspired by the mechanisms of the biological brain. Spiking neural networks (SNNs) provide a power-efficient and brain-inspired computing paradigm for machine learning applications. However, evaluating large-scale SNNs on classical von Neumann architectures (central processing units/graphics processing units) demands a high amount of power and time. Therefore, hardware designers have developed neuromorphic platforms to execute SNNs in and approach that combines fast processing and low power consumption. Recently, field-programmable gate arrays (FPGAs) have been considered promising candidates for implementing neuromorphic solutions due to their varied advantages, such as higher flexibility, shorter design, and excellent stability. This review aims to describe recent advances in SNNs and the neuromorphic hardware platforms (digital, analog, hybrid, and FPGA based) suitable for their implementation. We present that biological background of SNN learning, such as neuron models and information encoding techniques, followed by a categorization of SNN training. In addition, we describe state-of-the-art SNN simulators. Furthermore, we review and present FPGA-based hardware implementation of SNNs. Finally, we discuss some future directions for research in this field.

Neural readout of a latency code in the active electrosensory system

The latency of spikes relative to a stimulus conveys sensory information across modalities. However, in most cases, it remains unclear whether and how such latency codes are utilized by postsynaptic neurons. In the active electrosensory system of mormyrid fish, a latency code for stimulus amplitude in electroreceptor afferent nerve fibers (EAs) is hypothesized to be read out by a central reference provided by motor corollary discharge (CD). Here, we demonstrate that CD enhances sensory responses in postsynaptic granular cells of the electrosensory lobe but is not required for reading out EA input. Instead, diverse latency and spike count tuning across the EA population give rise to graded information about stimulus amplitude that can be read out by standard integration of converging excitatory synaptic inputs. Inhibitory control over the temporal window of integration renders two granular cell subclasses differentially sensitive to information derived from relative spike latency versus spike count.

Consciousness Beyond Neural Fields: Expanding the Possibilities of What Has Not Yet Happened

In the field theories in physics, any particular region of the presumed space-time continuum and all interactions between elementary objects therein can be objectively measured and/or accounted for mathematically. Since this does not apply to any of the field theories, or any other neural theory, of consciousness, their explanatory power is limited. As discussed in detail herein, the matter is complicated further by the facts than any scientifically operational definition of consciousness is inevitably partial, and that the phenomenon has no spatial dimensionality. Under the light of insights from research on meditation and expanded consciousness, chronic pain syndrome, healthy aging, and eudaimonic well-being, we may conceive consciousness as a source of potential energy that has no clearly defined spatial dimensionality, but can produce significant changes in others and in the world, observable in terms of changes in time. It is argued that consciousness may have evolved to enable the human species to generate such changes in order to cope with unprecedented and/or unpredictable adversity. Such coping could, ultimately, include the conscious planning of our own extinction when survival on the planet is no longer an acceptable option.

Analysis of the Information Capacity of Neuronal Molecular Communications under Demyelination and Remyelination

Demyelination of neurons can compromise the communication performance between the cells as the absence of myelin attenuates the action potential propagated through the axonal pathway. In this work, we propose a hybrid experimental and simulation model for analyzing the demyelination effects on neuron communication. The experiment involves locally induced demyelination using Lysolecithin and from this, a myelination index is empirically estimated from analysis of cell images. This index is then coupled with a modified Hodgkin-Huxley computational model to simulate the resulting impact that the de/myelination processes has on the signal propagation along the axon. The effects of signal degradation and transfer of neuronal information are simulated and quantified at multiple levels, and this includes (1) compartment per compartment of a single neuron, (2) bipartite synapse and the effects on the excitatory post-synaptic potential, and (3) a small network of neurons to understand how the impact of de/myelination has on the whole network. By using the myelination index in the simulation model, we can determine the level of attenuation of the action potential concerning the myelin quantity, as well as the analysis of internal signalling functions of the neurons and their impact on the overall spike firing rate. We believe that this hybrid experimental and in silico simulation model can result in a new analysis tool that can predict the gravity of the degeneration through the estimation of the spiking activity and vice-versa, which can minimize the need for specialised laboratory equipment needed for single-cell communication analysis.

A Double-Layer Multi-Resolution Classification Model for Decoding Spatiotemporal Patterns of Spikes With Small Sample Size

We build a double-layer, multiple temporal-resolution classification model for decoding single-trial spatiotemporal patterns of spikes. The model takes spiking activities as input signals and binary behavioral or cognitive variables as output signals and represents the input-output mapping with a double-layer ensemble classifier. In the first layer, to solve the underdetermined problem caused by the small sample size and the very high dimensionality of input signals, B-spline functional expansion and L1-regularized logistic classifiers are used to reduce dimensionality and yield sparse model estimations. A wide range of temporal resolutions of neural features is included by using a large number of classifiers with different numbers of B-spline knots. Each classifier serves as a base learner to classify spatiotemporal patterns into the probability of the output label with a single temporal resolution. A bootstrap aggregating strategy is used to reduce the estimation variances of these classifiers. In the second layer, another L1-regularized logistic classifier takes outputs of first-layer classifiers as inputs to generate the final output predictions. This classifier serves as a meta-learner that fuses multiple temporal resolutions to classify spatiotemporal patterns of spikes into binary output labels. We test this decoding model with both synthetic and experimental data recorded from rats and human subjects performing memory-dependent behavioral tasks. Results show that this method can effectively avoid overfitting and yield accurate prediction of output labels with small sample size. The double-layer, multi-resolution classifier consistently outperforms the best single-layer, single-resolution classifier by extracting and utilizing multi-resolution spatiotemporal features of spike patterns in the classification.

Supervised training of spiking neural networks for robust deployment on mixed-signal neuromorphic processors

Mixed-signal analog/digital circuits emulate spiking neurons and synapses with extremely high energy efficiency, an approach known as "neuromorphic engineering". However, analog circuits are sensitive to process-induced variation among transistors in a chip ("device mismatch"). For neuromorphic implementation of Spiking Neural Networks (SNNs), mismatch causes parameter variation between identically-configured neurons and synapses. Each chip exhibits a different distribution of neural parameters, causing deployed networks to respond differently between chips. Current solutions to mitigate mismatch based on per-chip calibration or on-chip learning entail increased design complexity, area and cost, making deployment of neuromorphic devices expensive and difficult. Here we present a supervised learning approach that produces SNNs with high robustness to mismatch and other common sources of noise. Our method trains SNNs to perform temporal classification tasks by mimicking a pre-trained dynamical system, using a local learning rule from non-linear control theory. We demonstrate our method on two tasks requiring temporal memory, and measure the robustness of our approach to several forms of noise and mismatch. We show that our approach is more robust than common alternatives for training SNNs. Our method provides robust deployment of pre-trained networks on mixed-signal neuromorphic hardware, without requiring per-device training or calibration.

Decoding neurobiological spike trains using recurrent neural networks: a case study with electrophysiological auditory cortex recordings

Recent advancements in multielectrode methods and spike-sorting algorithms enable the in vivo recording of the activities of many neurons at a high temporal resolution. These datasets offer new opportunities in the investigation of the biological neural code, including the direct testing of specific coding hypotheses, but they also reveal the limitations of present decoder algorithms. Classical methods rely on a manual feature extraction step, resulting in a feature vector, like the firing rates of an ensemble of neurons. In this paper, we present a recurrent neural-network-based decoder and evaluate its performance on experimental and artificial datasets. The experimental datasets were obtained by recording the auditory cortical responses of rats exposed to sound stimuli, while the artificial datasets represent preset encoding schemes. The task of the decoder was to classify the action potential timeseries according to the corresponding sound stimuli. It is illustrated that, depending on the coding scheme, the performance of the recurrent-network-based decoder can exceed the performance of the classical methods. We also show how randomized copies of the training datasets can be used to reveal the role of candidate spike-train features. We conclude that artificial neural network decoders can be a useful alternative to classical population vector-based techniques in studies of the biological neural code.

Integration, coincidence detection and resonance in networks of spiking neurons expressing Gamma oscillations and asynchronous states

Gamma oscillations are widely seen in the awake and sleeping cerebral cortex, but the exact role of these oscillations is still debated. Here, we used biophysical models to examine how Gamma oscillations may participate to the processing of afferent stimuli. We constructed conductance-based network models of Gamma oscillations, based on different cell types found in cerebral cortex. The models were adjusted to extracellular unit recordings in humans, where Gamma oscillations always coexist with the asynchronous firing mode. We considered three different mechanisms to generate Gamma, first a mechanism based on the interaction between pyramidal neurons and interneurons (PING), second a mechanism in which Gamma is generated by interneuron networks (ING) and third, a mechanism which relies on Gamma oscillations generated by pacemaker chattering neurons (CHING). We find that all three mechanisms generate features consistent with human recordings, but that the ING mechanism is most consistent with the firing rate change inside Gamma bursts seen in the human data. We next evaluated the responsiveness and resonant properties of these networks, contrasting Gamma oscillations with the asynchronous mode. We find that for both slowly-varying stimuli and precisely-timed stimuli, the responsiveness is generally lower during Gamma compared to asynchronous states, while resonant properties are similar around the Gamma band. We could not find conditions where Gamma oscillations were more responsive. We therefore predict that asynchronous states provide the highest responsiveness to external stimuli, while Gamma oscillations tend to overall diminish responsiveness.

In the awake and attentive brain, the activity of neurons is typically asynchronous and irregular. It also occasionally displays oscillations in the Gamma frequency range (30–90 Hz), which are believed to be involved in information processing. Here, we use computational models to investigate how brain circuits generate oscillations in a manner consistent with microelectrode recordings in humans. We then study how these networks respond to external input, comparing asynchronous and oscillatory states. This is tested according to several paradigms, an integrative mode, where slowly varying inputs are progressively integrated, a coincidence detection mode, where brief inputs are processed according to the phase of the oscillations, and a resonance mode where the network is probed with oscillatory inputs. Surprisingly, we find that in all cases, the presence of Gamma oscillations tends to diminish the responsiveness to external inputs. We discuss possible implications of this responsiveness decrease on information processing and propose new directions for further exploration.

A Survey of Encoding Techniques for Signal Processing in Spiking Neural Networks

Biologically inspired spiking neural networks are increasingly popular in the field of artificial intelligence due to their ability to solve complex problems while being power efficient. They do so by leveraging the timing of discrete spikes as main information carrier. Though, industrial applications are still lacking, partially because the question of how to encode incoming data into discrete spike events cannot be uniformly answered. In this paper, we summarise the signal encoding schemes presented in the literature and propose a uniform nomenclature to prevent the vague usage of ambiguous definitions. Therefore we survey both, the theoretical foundations as well as applications of the encoding schemes. This work provides a foundation in spiking signal encoding and gives an overview over different application-oriented implementations which utilise the schemes.

Illusory Essences: A Bias Holding Back Theorizing in Psychological Science

The reliance in psychology on verbal definitions means that psychological research is unusually moored to how humans think and communicate about categories. Psychological concepts (e.g., intelligence, attention) are easily assumed to represent objective, definable categories with an underlying essence. Like the “vital forces” previously thought to animate life, these assumed essences can create an illusion of understanding. By synthesizing a wide range of research lines from cognitive, clinical, and biological psychology and neuroscience, we describe a pervasive tendency across psychological science to assume that essences explain phenomena. Labeling a complex phenomenon can appear as theoretical progress before there is sufficient evidence that the described category has a definable essence or known boundary conditions. Category labels can further undermine progress by masking contingent and contextual relationships and obscuring the need to specify mechanisms. Finally, we highlight examples of promising methods that circumvent the lure of essences and suggest four concrete strategies for identifying and avoiding essentialist intuitions in theory development.

Integration of Leaky-Integrate-and-Fire Neurons in Standard Machine Learning Architectures to Generate Hybrid Networks: A Surrogate Gradient Approach

Up to now, modern machine learning (ML) has been based on approximating big data sets with high-dimensional functions, taking advantage of huge computational resources. We show that biologically inspired neuron models such as the leaky-integrate-and-fire (LIF) neuron provide novel and efficient ways of information processing. They can be integrated in machine learning models and are a potential target to improve ML performance. Thus, we have derived simple update rules for LIF units to numerically integrate the differential equations. We apply a surrogate gradient approach to train the LIF units via backpropagation. We demonstrate that tuning the leak term of the LIF neurons can be used to run the neurons in different operating modes, such as simple signal integrators or coincidence detectors. Furthermore, we show that the constant surrogate gradient, in combination with tuning the leak term of the LIF units, can be used to achieve the learning dynamics of more complex surrogate gradients. To prove the validity of our method, we applied it to established image data sets (the Oxford 102 flower data set, MNIST), implemented various network architectures, used several input data encodings and demonstrated that the method is suitable to achieve state-of-the-art classification performance. We provide our method as well as further surrogate gradient methods to train spiking neural networks via backpropagation as an open-source KERAS package to make it available to the neuroscience and machine learning community. To increase the interpretability of the underlying effects and thus make a small step toward opening the black box of machine learning, we provide interactive illustrations, with the possibility of systematically monitoring the effects of parameter changes on the learning characteristics.

Event-Based Trajectory Prediction Using Spiking Neural Networks

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.

Supervised Learning With First-to-Spike Decoding in Multilayer Spiking Neural Networks

Experimental studies support the notion of spike-based neuronal information processing in the brain, with neural circuits exhibiting a wide range of temporally-based coding strategies to rapidly and efficiently represent sensory stimuli. Accordingly, it would be desirable to apply spike-based computation to tackling real-world challenges, and in particular transferring such theory to neuromorphic systems for low-power embedded applications. Motivated by this, we propose a new supervised learning method that can train multilayer spiking neural networks to solve classification problems based on a rapid, first-to-spike decoding strategy. The proposed learning rule supports multiple spikes fired by stochastic hidden neurons, and yet is stable by relying on first-spike responses generated by a deterministic output layer. In addition to this, we also explore several distinct, spike-based encoding strategies in order to form compact representations of presented input data. We demonstrate the classification performance of the learning rule as applied to several benchmark datasets, including MNIST. The learning rule is capable of generalizing from the data, and is successful even when used with constrained network architectures containing few input and hidden layer neurons. Furthermore, we highlight a novel encoding strategy, termed "scanline encoding," that can transform image data into compact spatiotemporal patterns for subsequent network processing. Designing constrained, but optimized, network structures and performing input dimensionality reduction has strong implications for neuromorphic applications.

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?

Low-Latency Spiking Neural Networks Using Pre-Charged Membrane Potential and Delayed Evaluation

Spiking neural networks (SNNs) have attracted many researchers' interests due to its biological plausibility and event-driven characteristic. In particular, recently, many studies on high-performance SNNs comparable to the conventional analog-valued neural networks (ANNs) have been reported by converting weights trained from ANNs into SNNs. However, unlike ANNs, SNNs have an inherent latency that is required to reach the best performance because of differences in operations of neuron. In SNNs, not only spatial integration but also temporal integration exists, and the information is encoded by spike trains rather than values in ANNs. Therefore, it takes time to achieve a steady-state of the performance in SNNs. The latency is worse in deep networks and required to be reduced for the practical applications. In this work, we propose a pre-charged membrane potential (PCMP) for the latency reduction in SNN. A variety of neural network applications (e.g., classification, autoencoder using MNIST and CIFAR-10 datasets) are trained and converted to SNNs to demonstrate the effect of the proposed approach. The latency of SNNs is successfully reduced without accuracy loss. In addition, we propose a delayed evaluation method (DE), by which the errors during the initial transient are discarded. The error spikes occurring in the initial transient is removed by DE, resulting in the further latency reduction. DE can be used in combination with PCMP for further latency reduction. Finally, we also show the advantages of the proposed methods in improving the number of spikes required to reach a steady-state of the performance in SNNs for energy-efficient computing.

Time-Multiplexed Spiking Convolutional Neural Network Based on VCSELs for Unsupervised Image Classification

In this work, we present numerical results concerning a multilayer “deep” photonic spiking convolutional neural network, arranged so as to tackle a 2D image classification task. The spiking neurons used are typical two-section quantum-well vertical-cavity surface-emitting lasers that exhibit isomorphic behavior to biological neurons, such as integrate-and-fire excitability and timing encoding. The isomorphism of the proposed scheme to biological networks is extended by replicating the retina ganglion cell for contrast detection in the photonic domain and by utilizing unsupervised spike dependent plasticity as the main training technique. Finally, in this work we also investigate the possibility of exploiting the fast carrier dynamics of lasers so as to time-multiplex spatial information and reduce the number of physical neurons used in the convolutional layers by orders of magnitude. This last feature unlocks new possibilities, where neuron count and processing speed can be interchanged so as to meet the constraints of different applications.

Brain-inspired model for early vocal learning and correspondence matching using free-energy optimization

We propose a developmental model inspired by the cortico-basal system (CX-BG) for vocal learning in babies and for solving the correspondence mismatch problem they face when they hear unfamiliar voices, with different tones and pitches. This model is based on the neural architecture INFERNO standing for Iterative Free-Energy Optimization of Recurrent Neural Networks. Free-energy minimization is used for rapidly exploring, selecting and learning the optimal choices of actions to perform (eg sound production) in order to reproduce and control as accurately as possible the spike trains representing desired perceptions (eg sound categories). We detail in this paper the CX-BG system responsible for linking causally the sound and motor primitives at the order of a few milliseconds. Two experiments performed with a small and a large audio database show the capabilities of exploration, generalization and robustness to noise of our neural architecture in retrieving audio primitives during vocal learning and during acoustic matching with unheared voices (different genders and tones).

Differential Effect of Anesthesia on Visual Cortex Neurons with Diverse Population Coupling

Cortical neurons display diverse firing patterns and synchronization properties. How anesthesia alters the firing response of different neuron groups relevant for sensory information processing is unclear. Here we investigated the graded effect of anesthesia on spontaneous and visual flash-induced spike activity of different neuron groups classified based on their spike waveform, firing rate, and population coupling (the extent neurons conform to population spikes). Single-unit activity was measured from multichannel extracellular recordings in deep layers of primary visual cortex of freely moving rats in wakefulness and at three concentrations of desflurane. Anesthesia generally decreased firing rate and increased population coupling and burstiness of neurons. Population coupling and firing rate became more correlated and the pairwise correlation between neurons became more predictable by their population coupling in anesthesia. During wakefulness, visual stimulation increased firing rate; this effect was the largest and the most prolonged in neurons that exhibited high population coupling and high firing rate. During anesthesia, the early increase in firing rate (20-150 ms post-stimulus) of these neurons was suppressed, their spike timing was delayed and split into two peaks. The late response (200-400 ms post-stimulus) of all neurons was also suppressed. We conclude that anesthesia alters the visual response of primarily high-firing highly coupled neurons, which may interfere with visual sensory processing. The increased association of population coupling and firing rate during anesthesia suggests a decrease in sensory information content.