A Hybrid CMOS-Memristor Neuromorphic Synapse.

Area of Science:

• Neuroscience and Nano-electronics
• Neuromorphic Engineering
• Materials Science

Background:

• Current data processing struggles to replicate brain-like capabilities.
• Brain-inspired computing requires integrating neuroscience principles with advanced electronics.
• Existing memristive synapse designs have limitations in replicating complex biological behaviors.

Purpose of the Study:

• To propose a high-performance nano-scale Complementary Metal Oxide Semiconductor (CMOS)-memristive circuit.
• To mimic essential learning properties of biological synapses, specifically Spike Timing Dependent Plasticity (STDP).
• To advance neuromorphic architectures with improved synaptic functionalities.

Main Methods:

• Designed a hybrid circuit using memristors and CMOS transistors.
• Engineered the circuit to alter memristance based on pre- and post-synaptic action potential timing.
• Integrated the design into a crossbar array structure with silicon neurons.

Main Results:

• The proposed circuit effectively replicates essential behaviors observed in animal brain electrophysiology, including higher-order spike interactions.
• Achieved a significant reduction in CMOS area (factor of ten) compared to prior art.
• Demonstrated the circuit's ability to implement Spike Timing Dependent Plasticity (STDP).

Conclusions:

• The developed CMOS-memristive synaptic circuit offers a pathway to advanced neuromorphic systems.
• The design facilitates large-scale integration for applications in neuroscience research, pattern recognition, and Brain-Machine-Interfaces.
• This work represents a significant step towards realizing computers with brain-like processing and learning capabilities.