[Read Paper] Learning to Design Circuits
Learning to Design Circuits
Two difficult for searching for parameters that satisfy circuit specifications due to the low availability of training data:
- Circuit simulation is slow, thus generating large-scale dataset is time-consuming
- Most circuit designs are propitiatory IPs within individual IC companies, making it expensive to collect large-scale datasets
Constrains for L2DC(Learning to Design Circuits):
-
meet hard-constraints (eg. gain, bandwidth)
-
optimize good-to-have targets (eg. area, power)
Steps:
-
leverages reinforcement learning (RL) to generate circuits data by itself and learns from the data to search for best parameters.
- produces an action (a set of parameters) to the circuit simulator environment, and then receives a reward as a function of gain, bandwidth, power, area, etc.
- The reward is defined to optimize the desired Figures of Merits (FOM) composed of several performance metrics.
- By maximizing the reward, RL agent can optimize the circuit parameters.
