Array of Cointegrated Transistor-Based Artificial Neurons and Synapses for Neuromorphic Computing

Area of Science:

• Neuromorphic Engineering
• Integrated Circuits
• Artificial Intelligence Hardware

Background:

• Neuromorphic computing aims to mimic the brain's efficiency and parallelism.
• Existing systems face challenges with area, power consumption, and CMOS integration.

Purpose of the Study:

• To present a simplified, unified hardware architecture for neuromorphic computing.
• To overcome limitations of current circuit-based and non-CMOS neuromorphic systems.

Main Methods:

• Developed the Array of Cointegrated Transistor-Based Artificial Neurons and Synapses (ACTANS) architecture.
• Utilized homotypic transistors as functionally distinct artificial neurons and synapses.
• Implemented within a standard complementary metal-oxide-semiconductor (CMOS) fabrication process.

Main Results:

• ACTANS architecture offers a compact device footprint.
• Achieved seamless cointegration of neurons, synapses, and peripheral circuits.
• Demonstrated capability in performing cognitive tasks like letter and pattern recognition.

Conclusions:

• ACTANS provides a scalable and efficient solution for neuromorphic hardware.
• This approach facilitates the development of advanced AI systems with reduced resource demands.
• The unified transistor-based design represents a significant advancement in neuromorphic engineering.