ML Clustering for OPC Training Pattern Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for training machine learning models for wafer patterning processes are inefficient due to the manual selection of training patterns, which requires significant user expertise and time, often resulting in inadequate pattern coverage and inaccurate predictions, leading to lithography hotspots and process window limitations.
Innovation Solution
A method that generates a set of features from patterns in a pattern set, groups them based on similarities using machine learning clustering, and selects representative patterns to construct a training set for deep convolutional neural networks, ensuring adequate pattern coverage and improved prediction accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual selection of training patterns is used, then user expertise and control are required, but it consumes significant time and results in inadequate pattern coverage
Solution Approach 1:
The system performs self-service by automatically selecting training patterns through machine learning clustering algorithms without requiring manual user intervention. The automated pattern selector clusters patterns based on feature similarities and selects representative patterns, eliminating the need for user expertise while achieving adequate pattern coverage and improved prediction accuracy.
Solution Approach 2:
The manual mechanical process of user-driven pattern selection is replaced with an automated computational system using machine learning clustering algorithms. The system substitutes human expertise with automated feature extraction, clustering, and pattern selection mechanisms, significantly reducing time consumption while maintaining or improving prediction accuracy.
2Reliability
If manual selection of training patterns is used, then user control is maintained, but it leads to inadequate pattern coverage and inaccurate predictions
Solution Approach 1:
The system segments the pattern set into distinct clusters based on feature similarities using machine learning clustering algorithms. This segmentation allows the system to identify and select representative patterns from each cluster, ensuring comprehensive pattern coverage across different pattern types while maintaining high prediction accuracy through structured approach.
Solution Approach 2:
The manual control mechanism is replaced with automated machine learning-based pattern analysis that objectively evaluates pattern coverage and selects representative patterns. The system uses feature extraction and clustering algorithms to ensure adequate coverage of diverse pattern types, improving both reliability and adaptability simultaneously.
3Productivity
If automated pattern selection is implemented, then time consumption is reduced, but system complexity increases
Solution Approach 1:
The system extracts and isolates key features from patterns using machine learning techniques, separating essential pattern characteristics from redundant information. This extraction simplifies the subsequent clustering and selection processes, reducing overall system complexity while maintaining high productivity in automated pattern selection.
Solution Approach 2:
The system transforms patterns into feature representations through parameter changes, converting complex pattern data into simplified feature vectors suitable for clustering algorithms. This parameter transformation reduces computational complexity and enables efficient automated selection while improving productivity.
Data Source
AI summary
A method for determining a training pattern in a layout patterning process. The method includes generating a plurality of features from patterns in a pattern set; grouping the patterns in the pattern set into individual groups based on similarities in the plurality of generated features; and selecting representative patterns from the individual groups to determine the training pattern. In some embodiments, the method is a method for training a machine learning model in a layout patterning process. The method may include, for example, providing representative patterns from the individual groups to the machine learning model to train the machine learning model to predict a continuous transmission mask (CTM) map for optical proximity correction (OPC) in the layout patterning process.