Project description
Miniature material elements mimicking spiking neuron behaviour
The brain, a complex structure, integrates computing and memory through analogue signals transmitted among spiking neurons. Spiking Neural Networks (SNN) enable the development of neuromorphic computation systems that mimic biological brains. These systems excel at handling noisy data and stimuli, making them well-suited for perception, cognition, and motor tasks. However, replicating the behaviours of biological neurons requires advancements beyond existing design and manufacturing technologies. The ERC-funded PeroSpiker project aims to develop miniature material elements that emulate the behaviour of neurons and synapses, resulting in smaller, simpler, and more energy-efficient SNNs. The project focuses on metal halide perovskite, an ideal material for creating devices that mimic biological membranes and synapses.
Objective
A brain is a complex structure where computing and memory are tightly intertwined at very low power cost of operation, by analog signals across vast quantities of synapse-connected spiking neurons. Animal brains react intelligently to environmental events and perceptions. By developing similar Spiking Neural Networks (SNN) we can realize neuromorphic computation systems excellent for dealing with large amounts of noisy data and stimuli and very well suited for perception, cognition and motor tasks. But the current CMOS technologies perform very poorly for emulating the biological brains and their power consumption is large. Currently we cannot replicate biological neurons behaviours with existing design and manufacturing technology. This project aims to develop compact miniature material elements that will emulate closely the complex dynamic behaviour of neurons and synapses, to form SNNs with substantial reduction in footprint, complexity and energy cost for perception, learning and computation. We investigate the properties of metal halide perovskite that have produced excellent photovoltaic devices in the last decade. These perovskites have ionic/electronic conduction, hysteresis, memory effect and switchable and nonlinear behaviour, that make them ideally suited for the realization of devices in close fidelity to biological electrochemically gated membranes in neurons, and information-tracking synapses. We will use the methodology of impedance spectroscopy and equivalent circuit analysis to fabricate devices with dynamic responses emulating the natural neuronal coupling and synchronization. This method will produce the hardware that we need for a preferred spiking computational model, incorporating time, analog physical elements and dynamical complexity as computational tools. As illustration we will show visual object recognition from spiking data provided by a spiking retina by advanced neuristors and dynamic synapses.
Fields of science
- engineering and technologyelectrical engineering, electronic engineering, information engineeringelectronic engineeringanalogue electronics
- medical and health sciencesclinical medicineophthalmology
- natural sciencesphysical sciencesopticsspectroscopy
- natural sciencescomputer and information sciencesartificial intelligencecomputational intelligence
Programme(s)
- HORIZON.1.1 - European Research Council (ERC) Main Programme
Topic(s)
Funding Scheme
HORIZON-ERC - HORIZON ERC GrantsHost institution
46022 Valencia
Spain