A neuromorphic network for generic multivariate data classification

Abstract

Computational neuroscience has uncovered a number of computational principles used by nervous systems. At the same time, neuromorphic hardware has matured to a state where fast silicon implementations of complex neural networks have become feasible. En route to future technical applications of neuromorphic computing the current challenge lies in the identification and implementation of functional brain algorithms. Taking inspiration from the olfactory system of insects, we constructed a spiking neural network for the classification of multivariate data, a common problem in signal and data analysis. In this model, real-valued multivariate data are converted into spike trains using "virtual receptors" (VRs). Their output is processed by lateral inhibition and drives a winner-take-all circuit that supports supervised learning. VRs are conveniently implemented in software, whereas the lateral inhibition and classification stages run on accelerated neuromorphic hardware. When trained and tested on real-world datasets, we find that the classification performance is on par with a naïve Bayes classifier. An analysis of the network dynamics shows that stable decisions in output neuron populations are reached within less than 100 ms of biological time, matching the time-to-decision reported for the insect nervous system. Through leveraging a population code, the network tolerates the variability of neuronal transfer functions and trial-totrial variation that is inevitably present on the hardware system. Our work provides a proof of principle for the successful implementation of a functional spiking neural network on a configurable neuromorphic hardware system that can readily be applied to realworld computing problems.bioinspired computing | spiking networks | machine learning | multivariate classification T he remarkable sensory and behavioral capabilities of all higher organisms are provided by the network of neurons in their nervous systems. The computing principles of the brain have inspired many powerful algorithms for data processing, most importantly the perceptron and, building on top of that, multilayer artificial neural networks, which are being applied with great success to various data analysis problems (1). Although these networks operate with continuous values, computation in biological neuronal networks relies on the exchange of action potentials, or "spikes."Simulating networks of spiking neurons with software tools is computationally intensive, imposing limits to the duration of simulations and maximum network size. To overcome this limitation, several groups around the world have started to develop hardware realizations of spiking neuron models and neuronal networks (2-10) for studying the behavior of biological networks (11). The approach of the Spikey hardware system used in the present study is to enable high-throughput network simulations by speeding up computation by a factor of 10 4 compared with biological real time (12, 13). It has been developed as a reconfigurable multineuron computing substrat...

Predicting voluntary movements from motor cortical activity with neuromorphic hardware

“…This so-called lateral inhibition reduces input correlation and thereby improves classification performance [33], [36]. In the association layer there are two excitatory association neuron (ANe) populations that receive input from the projection neurons via plastic synapses.…”

“…The classifier spiking neural network used in this work modifies the network in [33]. Its threelayer architecture comprising an input layer, a decorrelation layer, and an association layer, is inspired by the insect olfactory system [36] (Figure 3).…”

“…Grouping neurons in small populations mitigates both these effects. The populations sizes that we adopted in this study have been chosen as a trade-off to minimize neuron count while counteracting these adverse effects as much as possible (see [33] and supplemental material therein for a detailed analysis).…”

“…First, since each of the 12 input populations that correspond to the 12 selected motor cortical cells consists of 6 neurons, we copied each of the 12 spike trains six times to supply all input neurons. Second, since simultaneous spikes would lead to undesired synchronous bursts in activity (see [33] and supplement therein for a detailed analysis), we introduced a small random delay, fixed for each spike train within a population, drawn from a uniform distribution with values between 0 ms and 100 ms. We wish to stress that the computational overhead on these operations is very low, essentially consisting of drawing one random number per input neuron, and executing 6 additions/delays per spike. Therefore these steps could easily be performed on live data in a realtime processing scenario, even with very modest low-power hardware.…”

“…SpiNNaker [24], NeuroGrid [25], ROLLS [26], IBM's TrueNorth [27], or the systems developed at the University of Heidelberg [28], [29] (see [30] for a review). Aside from testing principles of neural computation [28], [31], [32], initial applications of neuromorphic hardware have been demonstrated for generic pattern recognition [27], [33]. These applications highlight the potential of this new technology to solve various widely investigated computing problems.…”