V. Dante

A VLSI recurrent network of integrate-and-fire neurons connected by plastic synapses with long-term memory

Electronic neuromorphic devices with on-chip, on-line learning should be able to modify quickly the synaptic couplings to acquire information about new patterns to be stored (synaptic plasticity) and, at the same time, preserve this information on very long time scales (synaptic stability). Here, we illustrate the electronic implementation of a simple solution to this stability-plasticity problem, recently proposed and studied in various contexts. It is based on the observation that reducing the analog depth of the synapses to the extreme (bistable synapses) does not necessarily disrupt the performance of the device as an associative memory, provided that 1) the number of neurons is large enough; 2) the transitions between stable synaptic states are stochastic; and 3) learning is slow. The drastic reduction of the analog depth of the synaptic variable also makes this solution appealing from the point of view of electronic implementation and offers a simple methodological alternative to the technological solution based on floating gates. We describe the full custom analog very large-scale integration (VLSI) realization of a small network of integrate-and-fire neurons connected by bistable deterministic plastic synapses which can implement the idea of stochastic learning. In the absence of stimuli, the memory is preserved indefinitely. During the stimulation the synapse undergoes quick temporary changes through the activities of the pre- and postsynaptic neurons; those changes stochastically result in a long-term modification of the synaptic efficacy. The intentionally disordered pattern of connectivity allows the system to generate a randomness suited to drive the stochastic selection mechanism. We check by a suitable stimulation protocol that the stochastic synaptic plasticity produces the expected pattern of potentiation and depression in the electronic network.

A Multichip Pulse-Based Neuromorphic Infrastructure and Its Application to a Model of Orientation Selectivity

The growing interest in pulse-mode processing by neural networks is encouraging the development of hardware implementations of massively parallel networks of integrate-and-fire neurons distributed over multiple chips. Address-event representation (AER) has long been considered a convenient transmission protocol for spike based neuromorphic devices. One missing, long-needed feature of AER-based systems is the ability to acquire data from complex neuromorphic systems and to stimulate them using suitable data. We have implemented a general-purpose solution in the form of a peripheral component interconnect (PCI) board (the PCI-AER board) supported by software. We describe the main characteristics of the PCI-AER board, and of the related supporting software. To show the functionality of the PCI-AER infrastructure we demonstrate a reconfigurable multichip neuromorphic system for feature selectivity which models orientation tuning properties of cortical neurons

An aVLSI recurrent network of spiking neurons with reconfigurable and plastic synapses

We illustrate key features of an analog, VLSI (aVLSI) chip implementing a network composed of 32 integrate-and-fire (IF) neurons with firing rate adaptation (AHP current), endowed with both a recurrent synaptic connectivity and AER-based connectivity with external, AER-compliant devices. Synaptic connectivity can be reconfigured at will as for the presence/absence of each synaptic contact and the excitatory/inhibitory nature of each synapse. Excitatory synapses are plastic through a spike-driven stochastic, Hebbian mechanism, and possess a self-limiting mechanism aiming at an optimal use of synaptic resources for Hebbian learning

A VLSI network of spiking neurons with plastic fully configurable “stop-learning” synapses

We describe and demonstrate a neuromorphic, analog VLSI chip (termed F-LANN) hosting 128 integrate-and-fire (IF) neurons with spike-frequency adaptation, and 16,384 plastic bistable synapses implementing a self-regulated form of Hebbian, spike-driven, stochastic plasticity. The chip is designed to offer a high degree of reconfigurability: each synapse may be individually configured at any time to be either excitatory or inhibitory and to receive either recurrent input from an on-chip neuron or AER-based input from an off-chip neuron. The initial state of each synapse can be set as potentiated or depressed, and the state of each synapse can be read and stored on a computer.

A Neuromorphic aVLSI network chip with configurable plastic synapses

We describe and demonstrate the key features of a neuromorphic, analog VLSI chip (termed F-LANN) hosting 128 integrate-and-fire (IF) neurons with spike-frequency adaptation, and 16 384 plastic bistable synapses implementing a self-regulated form of Hebbian, spike-driven, stochastic plasticity. We were successfully able to test and verify the basic operation of the chip as well as its main new feature, namely the synaptic configurability. This configurability enables us to configure each individual synapse as either excitatory or inhibitory and to receive either recurrent input from an on-chip neuron or AER (address event representation)-based input from an off-chip neuron. Its also possible to set the initial state of each synapse as potentiated or depressed, and the state of each synapse can be read and stored on a computer. The main aim of this chip is to be able to efficiently perform associative learning experiments on a large number of synapses. In the future we would like to connect up multiple F-LANN chips together to be able to perform associative learning of natural stimulus sets.

A new dynamic tactile display for reconfigurable braille: implementation and tests

Different tactile interfaces have been proposed to represent either text (braille) or, in a few cases, tactile large-area screens as replacements for visual displays. None of the implementations so far can be customized to match users preferences, perceptual differences and skills. Optimal choices in these respects are still debated; we approach a solution by designing a flexible device allowing the user to choose key parameters of tactile transduction. We present here a new dynamic tactile display, a 8 × 8 matrix of plastic pins based on well-established and reliable piezoelectric technology to offer high resolution (pin gap 0.7mm) as well as tunable strength of the pins displacement, and refresh rate up to 50s−1. It can reproduce arbitrary patterns, allowing it to serve the dual purpose of providing, depending on contingent user needs, tactile rendering of non-character information, and reconfigurable braille rendering. Given the relevance of the latter functionality for the expected average user, we considered testing braille encoding by volunteers a benchmark of primary importance. Tests were performed to assess the acceptance and usability with minimal training, and to check whether the offered flexibility was indeed perceived by the subject as an added value compared to conventional braille devices. Different mappings between braille dots and actual tactile pins were implemented to match user needs. Performances of eight experienced braille readers were defined as the fraction of correct identifications of rendered content. Different information contents were tested (median performance on random strings, words, sentences identification was about 75%, 85%, 98%, respectively, with a significant increase, p < 0.01), obtaining statistically significant improvements in performance during the tests (p < 0.05). Experimental results, together with qualitative ratings provided by the subjects, show a good acceptance and the effectiveness of the proposed solution.

iTBS-induced LTP-like plasticity parallels oscillatory activity changes in the primary sensory and motor areas of macaque monkeys.

Recently, neuromodulation techniques based on the use of repetitive transcranial magnetic stimulation (rTMS) have been proposed as a non-invasive and efficient method to induce in vivo long-term potentiation (LTP)-like aftereffects. However, the exact impact of rTMS-induced perturbations on the dynamics of neuronal population activity is not well understood. Here, in two monkeys, we examine changes in the oscillatory activity of the sensorimotor cortex following an intermittent theta burst stimulation (iTBS) protocol. We first probed iTBS modulatory effects by testing the iTBS-induced facilitation of somatosensory evoked potentials (SEP). Then, we examined the frequency information of the electrocorticographic signal, obtained using a custom-made miniaturised multi-electrode array for electrocorticography, after real or sham iTBS. We observed that iTBS induced facilitation of SEPs and influenced spectral components of the signal, in both animals. The latter effect was more prominent on the θ band (4–8 Hz) and the high γ band (55–90 Hz), de-potentiated and potentiated respectively. We additionally found that the multi-electrode array uniformity of β (13–26 Hz) and high γ bands were also afflicted by iTBS. Our study suggests that enhanced cortical excitability promoted by iTBS parallels a dynamic reorganisation of the interested neural network. The effect in the γ band suggests a transient local modulation, possibly at the level of synaptic strength in interneurons. The effect in the θ band suggests the disruption of temporal coordination on larger spatial scales.

Median nerve stimulation modulates extracellular signals in the primary motor area of a macaque monkey.

Aiming to better define the functional influence of somatosensory stimuli on the primary motor cortex (M1) of primates, we investigated changes in extracellular neural activity induced by repetitive median nerve stimulation (MNS). We described neural adaptation and signal integration in both the multiunit activity (MUA) and the local field potential (LFP). To identify integration of initial M1 activity in the MNS response, we tested the correlation between peak amplitude responses and band energy preceding the peaks. Most of the sites studied in the M1 resulted responsive to MNS. MUA response peak amplitude decreased significantly in time in all sites during repetitive MNS, LFP response peak amplitude instead resulted more variable. Similarly, correlation analysis with the initial activity revealed a significant influence when tested using MUA peak amplitude modulation and a less significant correlation when tested using LFP peak amplitude. Our findings improve current knowledge on mechanisms underlying early M1 changes consequent to afferent somatosensory stimuli.

Interactive Reader Device for Visually Impaired People

In this feasibility study the design and development of an interactive reader device for blind is examined. It is designed around a character recognition engine based on artificial neural networks and has a modus operandi that elaborates a continuous stream of images, acquired by a custom hardware, at the same time the user moves the device on the text he is trying to read. Indeed the system has to guide the user to follow the text flow, in a suitable way. This makes the proposed solution interactive, since it is a complementary sensory aid that gives information about the text it reads (by means of a tactile display and a speech synthesis engine) and at the same time the user has the ability to select the right operating mode and the actual device configuration.