Method and apparatus for training model

Abstract

A method and an apparatus for training a model of a semiconductor device are provided. In the method, a training environment including a dataset is established, where the dataset correlates a plurality of parameters of the semiconductor device with at least one electrical parameter. A state represented by a set of values of the parameters is defined in the training environment. An action is executed through a prediction model to modify at least one value in the set of values. A reward value is calculated based on a reward function, where the reward function is configured to guide the prediction model to maximize the electrical parameter, subject to a parameter constraint. A violation of the parameter constraint is penalized by a penalty that increases non-linearly with a degree of the violation. The prediction model is updated based on the reward value. Therefore, design efficiency and accuracy could be significantly enhanced.

Images