Knowledge Distillation for Semiconductor Inspection Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current semiconductor inspection and metrology processes face challenges due to the need for multiple input modes and large training datasets, which can be costly and impractical, especially in terms of time and resource expenditure.
Innovation Solution
A system and method utilizing a knowledge distillation component to combine outputs from multiple deep learning models, followed by supervised training of a final knowledge distilled deep learning model, allowing for information determination in semiconductor inspection and metrology using fewer input modes and reducing computational requirements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple deep learning models with multiple input modes are used to improve detection accuracy, then measurement precision is improved, but device complexity and computational requirements increase
Solution Approach 1:
The patent extracts and distills essential knowledge from multiple complex deep learning models into a single simplified model. The knowledge distillation component captures the important patterns and relationships learned by the ensemble of models, then transfers this distilled knowledge to a reduced model that maintains high detection accuracy while significantly reducing computational complexity and the number of input modes required.
2Measurement precision
If large training datasets are used to train deep learning models, then measurement precision is improved, but loss of time and resources increases
Solution Approach 1:
The patent performs preliminary knowledge distillation during an initial phase, where multiple deep learning models are trained on available data and their essential knowledge is extracted and stored. This pre-computed distilled knowledge can then be used to train the final simplified model much faster without requiring re-processing of the entire original training dataset, significantly reducing the time and computational resources needed for subsequent detection tasks.
3Measurement precision
If multiple input modes are used to enhance model performance, then measurement precision is improved, but use of energy increases
Solution Approach 1:
The patent extracts and retains only the most essential and energy-efficient input modes that provide the maximum detection accuracy improvement. By identifying which input modes contribute most significantly to model performance during the knowledge distillation process, the system can eliminate redundant or less effective modes, thereby reducing the computational energy required for detection while maintaining high measurement precision.
Data Source
AI summary
Methods and systems for determining information for a specimen are provided. One system includes a computer subsystem and one or more components executed by the computer subsystem that include multiple deep learning (DL) models configured for determining information for a specimen based on output generated by the specimen with learning mode(s) of an imaging subsystem. The one or more components also include a knowledge distillation component configured for combining output generated by the multiple DL models. In addition, the one or more components include a final knowledge distilled DL model configured for determining information for the specimen or an additional specimen based on output generated for the specimen or the additional specimen with runtime mode(s) of the imaging subsystem. Before the final KD DL model determines the information, the knowledge distillation component is configured for supervised training of the final knowledge distilled DL model using the combined output.