Search

Spike-Timing-Dependent Plasticity (STDP) is a bio-inspired local incremental weight update rule commonly used for online learning in spike-based neuromorphic systems. In STDP, the intensity of long-term potentiation and depression in synaptic efficacy (weight) between neurons is expressed as a function of the relative timing between pre- and post-synaptic action potentials (spikes), while the polarity of change is dependent on the order (causality) of the spikes. Online STDP weight updates for causal and acausal relative spike times are activated at the onset of post- and pre-synaptic spike events, respectively, implying access to synaptic connectivity both in forward (pre-to-post) and reverse (post-to-pre) directions. Here we study the impact of different arrangements of synaptic connectivity tables on weight storage and STDP updates for large-scale neuromorphic systems. We analyze the memory efficiency for varying degrees of density in synaptic connectivity, ranging from crossbar arrays for full connectivity to pointer-based lookup for sparse connectivity. The study includes comparison of storage and access costs and efficiencies for each memory arrangement, along with a trade-off analysis of the benefits of each data structure depending on application requirements and budget. Finally, we present an alternative formulation of STDP via a delayed causal update mechanism that permits efficient weight access, requiring no more than forward connectivity lookup. We show functional equivalence of the delayed causal updates to the original STDP formulation, with substantial savings in storage and access costs and efficiencies for networks with sparse synaptic connectivity as typically encountered in large-scale models in computational neuroscience.

We demonstrate an event-driven Deep Learning (DL) hardware software ecosystem. The user-friendly software tools port models from Keras (popular machine learning libraries), automaticaly convert DL models to Spiking equivalents, i.e. Spiking Convolutional Neural Networks (SCNNs) and run spiking simulations of the converted models on the hardware emulator for testing and prototyping. More importantly, the software ports the converted models onto a novel, ultra-low power, real-time, event-driven ASIC SCNN Chip: DynapCNN. An interactive demonstration of a real-time face recognition system built using the above pipeline is shown as an example.

Identifying specific keywords in speech is a common task present in current mobile devices. In this context, Keyword spotting (KWS) is the triggering cue for the device to start “listening in” on the uttered command. Since this cue can originate at any moment, KWS must be continuously performed. Small-footprint KWS is typically designed to directly detect the entire keyword (as opposed to sub-word components, such as phonemes), with current low-power KWS solutions relying heavily on artificial neural networks (ANNs). As a power-efficient alternative to ANNs, biologically-inspired spiking neural networks (SNNs) have achieved similar results in terms of accuracy, with the benefit of simpler and, typically, sparser computation. SNNs can be designed to operate in rate- and time-based modes, each with its own particularities. In this work, we propose a small-footprint KWS solution on six common words from the TIMIT speech dataset using fully connected, feedforward neural networks, and compare the frame error rate performance and computational cost of ANN-versus-SNN alternatives. For the cost comparison, we designed a figure of merit which takes into account the types of operations present in digital implementations of the different types of networks. Lastly, we discuss the promising results for power-efficient KWS using SNNs, particularly in the time-based domain, which shows equivalent accuracy performance yet with an 84% reduction in cost in relation to its ANN equivalent.

While unsupervised generative neural networks are attractive choices for adoption in always-on continuous-time smart sensory systems, they typically impose heavy memory requirements on the underlying computational fabric. Recent literature on binarized neural networks has not yet been extended to unsupervised generative networks and alternate strategies are required to reduce their memory footprint. This work studies unsupervised synaptic pruning strategies to reduce the memory requirements for Restricted Boltzmann Machines (RBMs). In addition to one-shot pruning, we explore alternative strategies that encompass iterative stochastic pruning as well as pruning under target probability density functions for an RBM trained over the MNIST database. Interestingly, the results presented here suggest that one-shot re-training after pruning of the least significant connections in a trained network yields improved per-formance/memory trade-off over multiple iterations of stochastic pruning and re-training on the same network.

Metal oxide (MOX) gas sensors arrays are a predominant technological choice to perform fundamental tasks of chemical detection. Yet, their use has been mainly limited to relatively controlled instrument configurations where the sensor array is placed within a closed measurement chamber. Usually, the experimental protocol is defined beforehand and it includes three stages: the array is first exposed to a gas reference, then to the gas sample, and finally to the reference again to recover the initial state. Such sampling procedure requires signal acquisition during the complete experimental protocol and usually delays the output prediction until the predefined measurement duration is complete. Due to the slow time response of chemical sensors, the completion of the measurement typically requires minutes. In this paper we propose the use of reservoir computing (RC) algorithms to overcome the slow temporal dynamics of chemical sensor arrays, allowing identification and quantification of chemicals of interest continuously and reducing measurement delays. We generated two datasets to test the ability of RC algorithms to provide accurate and continuous prediction to fast varying gas concentrations in real time. Both datasets – one generated with synthetic data and the other acquired from actual gas sensors – provide time series of MOX sensors exposed to binary gas mixtures where concentration levels change randomly over time. Our results show that our approach improves the time response of the sensory system and provides accurate predictions in real time, making the system specifically suitable for online monitoring applications. Finally, the collected dataset and developed code are made publicly available to the research community for further studies.

Spike generation and routing is typically the most energy-demanding operation in neuromorphic hardware built using spiking neurons. Spiking neural networks running on neuromorphic hardware, however, often use rate-coding where the neurons spike rate is treated as the information-carrying quantity. Rate-coding is a highly inefficient coding scheme with minimal information content in each spike, which requires the transmission of a large number of spikes. In this paper, we describe an alternative type of spiking networks based on temporal coding where neuron spiking activity is very sparse and information is encoded in the time of each spike. We implemented the proposed networks on an FPGA platform and we use these sparsely active spiking networks to classify MNIST digits. The network FPGA implementation produces the classification output using only few tens of spikes from the hidden layer, and the classification result is obtained very quickly, typically within 1–3 synaptic time constants. We describe the idealized network dynamics and how these dynamics are adapted to allow an efficient implementation on digital hardware. Our results illustrate the importance of making use of the temporal dynamics in spiking networks in order to maximize the information content of each spike, which ultimately leads to reduced spike counts, improved energy efficiency, and faster response times.

In this paper we propose a method for continuously processing and learning from data in Restricted Boltzmann Machines (RBMs). Traditionally, RBMs are trained using Contrastive Divergence (CD), which is an algorithm consisting of two phases, of which only one is driven by data. This not only prohibits training of RBMs in conjugation with continuous-time data streams, especially in event-based real-time systems, but also hinders training speed of RBMs in large-scale machine learning systems. The model we propose trades space for time and, by pipelining information propagation in the network, is capable of processing both phases of the CD learning algorithm simultaneously. Simulation results of our model on generative and discriminative tasks show convergence to the original CD algorithm. We finalize with a discussion of applying our method to other deep neural networks, resulting in continuous learning and training time reduction.

Spiking neural networks are seen as the third generation of neural networks and the closest emulators of their biological counter parts. These networks use spikes as means of transmitting information between neurons. We study the merits and capacity of information transfer using spikes across different encoding and decoding schemes and show that spatio-temporal encoding scheme provides a very high efficiency in information transfer. We then explore learning rules based on neural dynamics that enable learning of spatio-temporal spike patterns. We explore various learning rules that can be used to learn spatio-temporal spike patterns.

Several learning rules for synaptic plasticity, that depend on either spike timing or internal state variables, have been proposed in the past imparting varying computational capabilities to Spiking Neural Networks. Due to design complications these learning rules are typically not implemented on neuromorphic devices leaving the devices to be only capable of inference. In this work we propose a unidirectional post-synaptic potential dependent learning rule that is only triggered by pre-synaptic spikes, and easy to implement on hardware. We demonstrate that such a learning rule is functionally capable of replicating computational capabilities of pairwise STDP. Further more, we demonstrate that this learning rule can be used to learn and classify spatio-temporal spike patterns in an unsupervised manner using individual neurons. We argue that this learning rule is computationally powerful and also ideal for hardware implementations due to its unidirectional memory access., Comment: Published in IEEE BioCAS 2016 Proceedings, BioMedical Circuits and Systems Conference, 2016

Brain plasticity serves animals in a wide range of vital functions. It assists them in adapting their behavior to the surroundings, in learning new strategies for optimizing a certain reward-seeking policy for their survival or in adjusting motor activity through sensory feedback. Thus, plasticity is an essential ingredient for building artificial autonomous systems that can cope with the real world. In order to build these systems, neuromorphic design labs actively investigate and develop various circuit implementations of plasticity. This research topic collects a comprehensive snapshot of this work. A number of manuscripts published in this topic study the interplay between stochasticity and plasticity (Afshar et al.; Bill and Legenstein; Lagorce et al.; Qiao et al.). Plasticity here acts in a stochastic fashion or extracts features from stochastic sensor data. The current push toward higher complexity/scale in neuromorphic devices can also be observed in plasticity implementations (Qiao et al.; Wang et al.; Noack et al.). Due to advantageous technology scaling and reproducibility, digital implementations of neuromorphic plasticity are gaining popularity (Galluppi et al.; Vogginger et al.). The collection of articles in this topic is rounded out by articles on plasticity in novel nano-scale technologies (Saighi et al.; Thomas et al.; Wang et al.; Bill and Legenstein).

Using a neural sampling approach, networks of stochastic spiking neurons, interconnected with plastic synapses, have been used to construct computational machines such as Restricted Boltzmann Machines (RBMs). Previous work towards building such networks achieved lower performances than traditional RBMs. More recently, Synaptic Sampling Machines (SSMs) were shown to outperform equivalent RBMs. In Synaptic Sampling Machines (SSMs), the stochasticity for the sampling is generated at the synapse. Stochastic synapses play the dual role of a regularizer during learning and an efficient mechanism for implementing stochasticity in neural networks over a wide dynamic range. In this paper we show that SSMs with stochastic synapses implemented in FPGA-based spiking neural networks can obtain a high accuracy in classifying MNIST handwritten digit database. We compare classification accuracy for different bit precision for stochastic and non-stochastic synapses and further argue that stochastic synapses have the same effect as synapses with higher bit precision but require significantly lower computational resources.

Spike-timing-dependent plasticity (STDP) incurs both causal and acausal synaptic weight updates, for negative and positive time differences between pre-synaptic and post-synaptic spike events. For realizing such updates in neuromorphic hardware, current implementations either require forward and reverse lookup access to the synaptic connectivity table, or rely on memory-intensive architectures such as crossbar arrays. We present a novel method for realizing both causal and acausal weight updates using only forward lookup access of the synaptic connectivity table, permitting memory-efficient implementation. A simplified implementation in FPGA, using a single timer variable for each neuron, closely approximates exact STDP cumulative weight updates for neuron refractory periods greater than 10 ms, and reduces to exact STDP for refractory periods greater than the STDP time window. Compared to conventional crossbar implementation, the forward table-based implementation leads to substantial memory savings for sparsely connected networks supporting scalable neuromorphic systems with fully reconfigurable synaptic connectivity and plasticity., Comment: Submitted to BioCAS 2016

Neuromorphic hardware offers an electronic substrate for the realization of asynchronous event-based sensory-motor systems and large-scale spiking neural network architectures. In order to characterize these systems, configure them, and carry out modeling experiments, it is often necessary to interface them to workstations. The software used for this purpose typically consists of a large monolithic block of code which is highly specific to the hardware setup used. While this approach can lead to highly integrated hardware/software systems, it hampers the development of modular and reconfigurable infrastructures thus preventing a rapid evolution of such systems. To alleviate this problem, we propose PyNCS, an open-source front-end for the definition of neural network models that is interfaced to the hardware through a set of Python Application Programming Interfaces (APIs). The design of PyNCS promotes modularity, portability and expandability and separates implementation from hardware description. The high-level front-end that comes with PyNCS includes tools to define neural network models as well as to create, monitor and analyze spiking data. Here we report the design philosophy behind the PyNCS framework and describe its implementation. We demonstrate its functionality with two representative case studies, one using an event-based neuromorphic vision sensor, and one using a set of multi-neuron devices for carrying out a cognitive decision-making task involving state-dependent computation. PyNCS, already applicable to a wide range of existing spike-based neuromorphic setups, will accelerate the development of hybrid software/hardware neuromorphic systems, thanks to its code flexibility. The code is open-source and available online at https://github.com/inincs/pyNCS., Frontiers in Neuroinformatics, 8, ISSN:1662-5196

We have recently demonstrated the emergence of dynamic feature sensitivity through exposure to formative stimuli in a real-time neuromorphic system implementing a hybrid analogue/digital network of spiking neurons. This network, inspired by models of auditory processing in mammals, includes several mutually connected layers with distance-dependent transmission delays and learning in the form of spike timing dependent plasticity, which effects stimulus-driven changes in the network connectivity.Here we present results that demonstrate that the network is robust to a range of variations in the stimulus pattern, such as are found in naturalistic stimuli and neural responses. This robustness is a property critical to the development of realistic, electronic neuromorphic systems.We analyse the variability of the response of the network to `noisy' stimuli which allows us to characterize the acuity in information-theoretic terms. This provides an objective basis for the quantitative comparison of networks, their connectivity patterns, and learning strategies, which can inform future design decisions. We also show, using stimuli derived from speech samples, that the principles are robust to other challenges, such as variable presentation rate, that would have to be met by systems deployed in the real world. Finally we demonstrate the potential applicability of the approach to real sounds.

Stimulus-specific adaptation (SSA) is a phenomenon observed in neural systems which occurs when the spike count elicited in a single neuron decreases with repetitions of the same stimulus, and recovers when a different stimulus is presented. SSA therefore effectively highlights rare events in stimulus sequences, and suppresses responses to repetitive ones. In this paper we present a model of SSA based on synaptic depression and describe its implementation in neuromorphic analog very-large-scale integration (VLSI). The hardware system is evaluated using biologically realistic spike trains with parameters chosen to reflect those of the stimuli used in physiological experiments. We examine the effect of input parameters and stimulus history upon SSA and show that the trends apparent in the results obtained in silico compare favorably with those observed in biological neurons.

Spike-based neuromorphic electronic architectures offer an attractive solution for implementing compact efficient sensory-motor neural processing systems for robotic applications. Such systems typically comprise event-based sensors and multi-neuron chips that encode, transmit, and process signals using spikes. For robotic applications, the ability to sustain real-time interactions with the environment is an essential requirement. So these neuromorphic systems need to process sensory signals continuously and instantaneously, as the input data arrives, classify the spatio-temporal information contained in the data, and produce appropriate motor outputs in real-time. In this paper we evaluate the computational approaches that have been proposed for classifying spatio-temporal sequences of spike-trains, derive the main principles and the key components that are required to build a neuromorphic system that works in robotic application scenarios, with the constraints imposed by the biologically realistic hardware implementation, and present possible system-level solutions.

Axonal delays are used in neural computation to implement faithful models of biological neural systems, and in spiking neural networks models to solve computationally demanding tasks. While there is an increasing number of software simulations of spiking neural networks that make use of axonal delays, only a small fraction of currently existing hardware neuromorphic systems supports them. In this paper we demonstrate a strategy to implement temporal delays in hardware spiking neural networks distributed across multiple Very Large Scale Integration (VLSI) chips. This is achieved by exploiting the inherent device mismatch present in the analog circuits that implement silicon neurons and synapses inside the chips, and the digital communication infrastructure used to configure the network topology and transmit the spikes across chips. We present an example of a recurrent VLSI spiking neural network that employs axonal delays and demonstrate how the proposed strategy efficiently implements them in hardware.

Many sounds of ecological importance, such as communication calls, are characterized by time-varying spectra. However, most neuromorphic auditory models to date have focused on distinguishing mainly static patterns, under the assumption that dynamic patterns can be learned as sequences of static ones. In contrast, the emergence of dynamic feature sensitivity through exposure to formative stimuli has been recently modeled in a network of spiking neurons based on the thalamo-cortical architecture. The proposed network models the effect of lateral and recurrent connections between cortical layers, distance-dependent axonal transmission delays, and learning in the form of Spike Timing Dependent Plasticity (STDP), which effects stimulus-driven changes in the pattern of network connectivity. In this paper we demonstrate how these principles can be efficiently implemented in neuromorphic hardware. In doing so we address two principle problems in the design of neuromorphic systems: real-time event-based asynchronous communication in multi-chip systems, and the realization in hybrid analog/digital VLSI technology of neural computational principles that we propose underlie plasticity in neural processing of dynamic stimuli. The result is a hardware neural network that learns in real-time and shows preferential responses, after exposure, to stimuli exhibiting particular spectro-temporal patterns. The availability of hardware on which the model can be implemented, makes this a significant step toward the development of adaptive, neurobiologically plausible, spike-based, artificial sensory systems., Frontiers in Neuroscience, 6, ISSN:1662-453X, ISSN:1662-4548

In the past recent years several research groups have proposed neuromorphic Very Large Scale Integration (VLSI) devices that implement event-based sensors or biophysically realistic networks of spiking neurons. It has been argued that these devices can be used to build event-based systems, for solving real-world applications in real-time, with efficiencies and robustness that cannot be achieved with conventional computing technologies. In order to implement complex event-based neuromorphic systems it is necessary to interface the neuromorphic VLSI sensors and devices among each other, to robotic platforms, and to workstations (e.g. for data-logging and analysis). This apparently simple goal requires painstaking work that spans multiple levels of complexity and disciplines: from the custom layout of microelectronic circuits and asynchronous printed circuit boards, to the development of object oriented classes and methods in software; from electrical engineering and physics for analog/digital circuit design to neuroscience and computer science for neural computation and spike-based learning methods. Within this context, we present a framework we developed to simplify the configuration of multi-chip neuromorphic VLSI systems, and automate the mapping of neural network model parameters to neuromorphic circuit bias values.

Stimulus-specific adaptation (SSA) is a phenomenon observed in neural systems which occurs when the spike count elicited in a single neuron by external stimuli decreases with repetitions of the same stimulus, and recovers when a different stimulus is presented. SSA therefore effectively highlights rare events in stimulus sequences, and suppresses responses to repetitive ones. In this paper we present a model of SSA based on synaptic depression and describe its implementation in neuromorphic analog VLSI. The hardware system is evaluated using biologically realistic spike trains with parameters chosen to match those used in physiological experiments. We examine the effect of input parameters upon SSA and show that the trends apparent in the results obtained in silico compare favourably with those observed in biological neurons.

Chemoresistive gas sensors, and Metal Oxide (MOX) sensors in particular, are predominant choices to perform funda- mental tasks of chemical detection due to their sensitivity, ease of use, low-cost, and fast response time compared to other technologies. Yet, their use has been mainly limited to relatively controlled scenarios where a gas sensor array is first exposed to a reference, then to the gas sample, and finally to the reference again to recover the initial state. In this paper we propose the use of MOX sensors along with Reservoir Computing algorithms to identify chemicals of interest. Our approach allows continuous gas monitoring in simple experimental setups without the requirement of acquiring recovery transient of the sensors, thereby making the system specifically suitable for online monitoring applications.