The NeuroModeler Laboratory

Advanced neuro-modeling techniques for computer-aided design of high-frequency electronic circuits are developed to exploit the power of neural networks. Fast neural network based models are developed for high-frequency components, both linear and nonlinear, using EM-based training for passive component models and physics-based training for active device models. Techniques for circuit design and yield optimization using passive/active neural models are also being developed. A recent breakthrough is a dynamic neural network based technique for behavioural modeling of nonlinear circuits directly from external data. The work will be useful for modeling and design optimization of highspeed, high frequency IC packages and receiver/transmitter circuits.