Accelerated training of machine learning based models for semiconductor applications

摘要

Methods and systems are provided for accelerated training of machine learning based models for semiconductor applications. One way to train a machine learning based model is to obtain information for non-nominal instances of a sample, which process is performed based on this information, which is to be obtained . A machine learning based model is configured to perform a simulation for the sample. Machine learning based models are trained with only information for nominal instances of additional samples. The method also retrains the machine learning based model with the information for the non-nominal instance of the sample, thereby to the machine learning based model of the information for the non-nominal instance of the sample. Includes performing transfer learning.