A Model of the Early Visual System Based on Parallel Spike-Sequence Detection, Showing Orientation Selectivity

Simple Summary

A computational model of primates’ early visual processing, showing orientation selectivity, is presented. The system importantly integrates two key elements: (1) a neuromorphic spike-decoding structure that considerably resembles the circuitry between layers IV and II/III of the primary visual cortex, both in topology and operation; (2) the plasticity of intrinsic excitability, to embed recent findings about the operation of the same area. The model is proposed as a tool for the analysis and reproduction of the orientation selectivity phenomenon, whose underlying neuronal-level computational mechanisms are today the subject of intense scrutiny. In response to rotated Gabor patches the model is able to exhibit realistic orientation tuning curves and to reproduce responses similar to those found in neurophysiological recordings from the primary visual cortex obtained under the same task, considering different stages of the network. This demonstrates its aptness to capture the mechanisms underlying the evoked response in the primary visual cortex. Our tool is available online, and can be expanded to other experiments using a dedicated software library developed by the authors, to elucidate the computational mechanisms underlying orientation selectivity.

Abstract

Since the first half of the twentieth century, numerous studies have been conducted on how the visual cortex encodes basic image features. One of the hallmarks of basic feature extraction is the phenomenon of orientation selectivity, of which the underlying neuronal-level computational mechanisms remain partially unclear despite being intensively investigated. In this work we present a reduced visual system model (RVSM) of the first level of scene analysis, involving the retina, the lateral geniculate nucleus and the primary visual cortex (V1), showing orientation selectivity. The detection core of the RVSM is the neuromorphic spike-decoding structure MNSD, which is able to learn and recognize parallel spike sequences and considerably resembles the neuronal microcircuits of V1 in both topology and operation. This structure is equipped with plasticity of intrinsic excitability to embed recent findings about V1 operation. The RVSM, which embeds 81 groups of MNSD arranged in 4 oriented columns, is tested using sets of rotated Gabor patches as input. Finally, synthetic visual evoked activity generated by the RVSM is compared with real neurophysiological signal from V1 area: (1) postsynaptic activity of human subjects obtained by magnetoencephalography and (2) spiking activity of macaques obtained by multi-tetrode arrays. The system is implemented using the NEST simulator. The results attest to a good level of resemblance between the model response and real neurophysiological recordings. As the RVSM is available online, and the model parameters can be customized by the user, we propose it as a tool to elucidate the computational mechanisms underlying orientation selectivity.

Temporal Backpropagation for Spiking Neural Networks with One Spike per Neuron

We propose a new supervised learning rule for multilayer spiking neural networks (SNNs) that use a form of temporal coding known as rank-order-coding. With this coding scheme, all neurons fire exactly one spike per stimulus, but the firing order carries information. In particular, in the readout layer, the first neuron to fire determines the class of the stimulus. We derive a new learning rule for this sort of network, named S4NN, akin to traditional error backpropagation, yet based on latencies. We show how approximated error gradients can be computed backward in a feedforward network with any number of layers. This approach reaches state-of-the-art performance with supervised multi-fully connected layer SNNs: test accuracy of 97.4% for the MNIST dataset, and 99.2% for the Caltech Face/Motorbike dataset. Yet, the neuron model that we use, nonleaky integrate-and-fire, is much simpler than the one used in all previous works. The source codes of the proposed S4NN are publicly available at https://github.com/SRKH/S4NN .

Category Structure and Categorical Perception Jointly Explained by Similarity-Based Information Theory

Categorization is a fundamental information processing phenomenon in the brain. It is critical for animals to compress an abundance of stimulations into groups to react quickly and efficiently. In addition to labels, categories possess an internal structure: the goodness measures how well any element belongs to a category. Interestingly, this categorization leads to an altered perception referred to as categorical perception: for a given physical distance, items within a category are perceived closer than items in two different categories. A subtler effect is the perceptual magnet: discriminability is reduced close to the prototypes of a category and increased near its boundaries. Here, starting from predefined abstract categories, we naturally derive the internal structure of categories and the phenomenon of categorical perception, using an information theoretical framework that involves both probabilities and pairwise similarities between items. Essentially, we suggest that pairwise similarities between items are to be tuned to render some predefined categories as well as possible. However, constraints on these pairwise similarities only produce an approximate matching, which explains concurrently the notion of goodness and the warping of perception. Overall, we demonstrate that similarity-based information theory may offer a global and unified principled understanding of categorization and categorical perception simultaneously.

WITHDRAWN: Auditory Surprise Model Based on Pattern Retrieval from the Past Observation

This article has been withdrawn: please see Elsevier Policy on Article Withdrawal (http://www.elsevier.com/locate/withdrawalpolicy). This article has been withdrawn at the request of the authors. The authors regrets that the reason for withdrawal is due to an disagreement in authorship and in scope of data disclosure. The authors apologize to the readers for this unfortunate error.

Neuromorphic Artificial Touch for Categorization of Naturalistic Textures

We implemented neuromorphic artificial touch and emulated the firing behavior of mechanoreceptors by injecting the raw outputs of a biomimetic tactile sensor into an Izhikevich neuronal model. Naturalistic textures were evaluated with a passive touch protocol. The resulting neuromorphic spike trains were able to classify ten naturalistic textures ranging from textiles to glass to BioSkin, with accuracy as high as 97%. Remarkably, rather than on firing rate features calculated over the stimulation window, the highest achieved decoding performance was based on the precise spike timing of the neuromorphic output as captured by Victor Purpura distance. We also systematically varied the sliding velocity and the contact force to investigate the role of sensing conditions in categorizing the stimuli via the artificial sensory system. We found that the decoding performance based on the timing of neuromorphic spike events was robust for a broad range of sensing conditions. Being able to categorize naturalistic textures in different sensing conditions, these neurorobotic results pave the way to the use of neuromorphic tactile sensors in future real-life neuroprosthetic applications.

FROM SENSORS TO SPIKES: EVOLVING RECEPTIVE FIELDS TO ENHANCE SENSORIMOTOR INFORMATION IN A ROBOT-ARM

In biological systems, instead of actual encoders at different joints, proprioception signals are acquired through distributed receptive fields. In robotics, a single and accurate sensor output per link (encoder) is commonly used to track the position and the velocity. Interfacing bio-inspired control systems with spiking neural networks emulating the cerebellum with conventional robots is not a straight forward task. Therefore, it is necessary to adapt this one-dimensional measure (encoder output) into a multidimensional space (inputs for a spiking neural network) to connect, for instance, the spiking cerebellar architecture; i.e. a translation from an analog space into a distributed population coding in terms of spikes. This paper analyzes how evolved receptive fields (optimized towards information transmission) can efficiently generate a sensorimotor representation that facilitates its discrimination from other "sensorimotor states". This can be seen as an abstraction of the Cuneate Nucleus (CN) functionality in a robot-arm scenario. We model the CN as a spiking neuron population coding in time according to the response of mechanoreceptors during a multi-joint movement in a robot joint space. An encoding scheme that takes into account the relative spiking time of the signals propagating from peripheral nerve fibers to second-order somatosensory neurons is proposed. Due to the enormous number of possible encodings, we have applied an evolutionary algorithm to evolve the sensory receptive field representation from random to optimized encoding. Following the nature-inspired analogy, evolved configurations have shown to outperform simple hand-tuned configurations and other homogenized configurations based on the solution provided by the optimization engine (evolutionary algorithm). We have used artificial evolutionary engines as the optimization tool to circumvent nonlinearity responses in receptive fields.

Encoding/decoding of first and second order tactile afferents in a neurorobotic application

We present a neurorobotic framework to investigate tactile information processing at the early stages of the somatosensory pathway. We focus on spatiotemporal coding of first and second order responses to Braille stimulation, which offers a suitable protocol to investigate the neural bases of fine touch discrimination. First, we model Slow Adaptive type I fingertip mechanoreceptor responses to Braille characters sensed both statically and dynamically. We employ a network of spiking neurones to transduce analogue skin deformations into primary spike trains. Then, we model second order neurones in the cuneate nucleus (CN) of the brainstem to study how mechanoreceptor responses are possibly processed prior to their transmission to downstream central areas. In the model, the connectivity layout of mechanoreceptor-to-cuneate projections produces a sparse CN code. To characterise the reliability of neurotransmission we employ an information theoretical measure accounting for the metrical properties of spiking signals. Our results show that perfect discrimination of primary and secondary responses to a set of 26 Braille characters is achieved within 100 and 500 ms of stimulus onset, in static and dynamic conditions, respectively. Furthermore, clusters of responses to different stimuli are better separable after the CN processing. This finding holds for both statically and dynamically delivered stimuli. In the presented system, when sliding the artificial fingertip over a Braille line, a speed of 40-50mm/s is optimal in terms of rapid and reliable character discrimination. This result is coherent with psychophysical observations reporting average reading speeds of 30-40±5 mm/s adopted by expert Braille readers.