Tunable low energy, compact and high performance neuromorphic circuit for spike-based synaptic plasticity

Area of Science:

• Neuromorphic Engineering
• Computational Neuroscience
• Analog Circuit Design

Background:

• Cortical circuits' information processing is studied via Spiking Neural Networks (SNNs).
• Neuromorphic engineering aims to replicate biological computation in silico for low power and compactness.
• Existing neuromorphic circuits face challenges in mimicking biological dynamics and stability.

Purpose of the Study:

• To propose a new accelerated-time circuit for Spiking Neural Networks (SNNs).
• To demonstrate advantages in compactness, power consumption, and biological mimicry over existing designs.
• To evaluate the circuit's stability against process variations and transistor mismatch.

Main Methods:

• Circuit design and simulation of an accelerated-time neuromorphic circuit.
• Comparison with existing synaptic plasticity circuits based on silicon area and energy consumption.
• Monte Carlo simulations to assess tolerance to mismatch and process variation.

Main Results:

• The proposed circuit achieves reduced silicon area and lower energy consumption per spike.
• The circuit effectively mimics spike timing- and rate-based synaptic plasticity experiments.
• Monte Carlo results show superior stability against transistor mismatch compared to prior designs.

Conclusions:

• The new circuit offers significant improvements in efficiency and stability for neuromorphic applications.
• Its ability to mimic biological plasticity makes it suitable for advanced learning systems.
• The design is ideal for large-scale SNNs requiring adaptability and computational power.