The objective of this research is to develop new neurodynamic programming (NDP) learning algorithm for controlling neuron-level activity (spiking) and synaptic-level plasticity in CMOS/memristor devices, such that the subsequent system-level response achieves desired sensorimotor behavioral goals. The approach is to uses a radically new training paradigm that induces functional plasticity by controlling the neural activity of selected input neurons via programming voltages, rather than by directly manipulating the synaptic weights, as do virtually all existing training algorithms.
Intellectual merit This research aims to develop a model of the closures required to translate synaptic-level plasticity into functional-level plasticity that results into high-level behavioral goals and problem solving abilities. The same critical challenge has been identified in neuroscience research aimed at reverse engineering the brain, and in the regulation of deep-brain stimulation (DBS). Due to this knowledge gap, even when a measure of adequate or desired behavior is available, it may not be easily utilized to stimulate a neural network at the cell level in order to produce the appropriate macroscopic behavior.
Broader impact The learning model developed in this research will be used toward the development of nanoscale neuromorphic systems that mimic neuro-biological architectures in the nervous system. Thanks to their abilities to recreate the synaptic plasticity, device density, scalability, and fault-tolerance of biological neuronal networks, these neuromorphic systems can enable a wide range of technological advancements, such as intelligent robots with highly-sophisticated sensorimotor skills, and neuroprosthetic devices capable of adapting to changing conditions and environments.
