Implementation of a spike-based perceptron learning rule using TiO2-x memristors

Area of Science:

• Neuroscience
• Materials Science
• Computer Engineering

Background:

• Synaptic plasticity is fundamental for neural network learning and adaptation.
• Neuromorphic systems require plastic synapses for cognitive functions like learning.
• Nano-scale memristors offer a scalable solution for implementing artificial synapses.

Purpose of the Study:

• To propose and demonstrate a hybrid CMOS-memristor system for neuromorphic learning.
• To implement and evaluate a spike-based Perceptron plasticity rule for synaptic modulation.
• To showcase the advantages of this rule over traditional STDP rules.

Main Methods:

• Development of a hybrid system integrating CMOS neurons with TiO2-x memristors.
• Implementation of spike-based learning circuits to control memristor conductance.
• Experimental validation using two silicon neurons and a memristive synapse.

Main Results:

• Demonstrated stable changes in memristor conductance induced by CMOS plasticity circuits.
• Observed increased synaptic strength after potentiation episodes.
• Observed decreased synaptic strength after depression episodes.

Conclusions:

• The proposed hybrid CMOS-memristor system effectively implements synaptic plasticity.
• The spike-based Perceptron rule enables stable and adaptive learning in neuromorphic circuits.
• This approach shows promise for developing advanced cognitive neuromorphic systems.