VLSI circuits implementing computational models of neocortical circuits.

Area of Science:

• Neuroscience
• Computer Engineering
• Artificial Intelligence

Background:

• Neuromorphic computing aims to emulate brain functions for advanced cognitive systems.
• Existing computational models often face limitations in speed and power efficiency.

Purpose of the Study:

• To design and implement novel neuromorphic integrated circuits for the COLAMN project.
• To enable accelerated, low-power execution of complex nonlinear neural models.

Main Methods:

• Developed three neuromorphic integrated circuits using 0.35 μm CMOS technology.
• Incorporated spiking/bursting neuron models and various synapse dynamics (short-term and long-term plasticity).
• Utilized analogue techniques for neural dynamics and asynchronous digital event-based I/O.

Main Results:

• Fabricated circuits demonstrate the feasibility of analogue implementation for neural dynamics.
• Circuits support configurable hardware blocks for diverse neural network simulations.
• Experimental results validate the performance of the designed neuromorphic circuits.

Conclusions:

• Analogue circuit techniques offer a viable approach for low-power, high-speed neuromorphic computing.
• The COLAMN project's circuits provide a flexible platform for cognitive system research.
• Further exploration of analogue vs. digital trade-offs is warranted for future neuromorphic designs.