Neural dynamics in reconfigurable silicon.

Area of Science:

• Neuroscience
• Computer Engineering
• Artificial Intelligence

Background:

• Neuromorphic engineering seeks to mimic brain function in hardware.
• Existing systems often lack the programmability for detailed neural dynamics simulation.
• Efficient implementation of complex neural models remains a challenge.

Purpose of the Study:

• To present a novel neuromorphic analog chip with high computational efficiency and programmability.
• To demonstrate the chip's capability in implementing diverse neural components and models.
• To enable advanced simulations and direct interfacing with biological systems.

Main Methods:

• Design and fabrication of a neuromorphic analog chip with 28 Computational Analog Blocks (CABs).
• Implementation of various neural components including neurons (Hopf bifurcations, integrate-and-fire), synapses, and central pattern generators.
• Utilization of floating gate transistors for on-chip programmability and reconfigurable interconnects.

Main Results:

• Successful implementation and measurement of detailed neuron dynamics and neural network components.
• Demonstration of massive computational area efficiency with over 50,000 synapses in 9 mm².
• Validation of the chip as a platform for both detailed simulations and dynamic clamp applications.

Conclusions:

• The presented neuromorphic chip offers a powerful and efficient platform for simulating complex neural dynamics.
• Its programmability and high synaptic density pave the way for advanced brain-inspired computing.
• The chip facilitates novel research at the intersection of computational neuroscience and bio-interfacing.