Substrate Processing Prediction With Spatial Feature Dimension Mapping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning models for substrate processing lack accuracy and interpretability due to the failure to consider spatial correlation, leading to distorted predictions and difficulty in identifying effective parameters at specific sites.
Innovation Solution
Introduce dimension mapping into a machine learning model to account for spatial correlation, using a feature extraction model to convert feature dimensions and a prediction model to compute predicted values, enhancing accuracy and interpretability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If spatial correlation is not considered in the learning model, then the model complexity remains low, but the prediction accuracy deteriorates due to distorted predictions
Solution Approach 1:
The patent applies dimensionality change by converting feature dimensions through dimension mapping to reflect spatial correlation. The feature conversion unit transforms input features into converted features with different dimensions that encode spatial relationships, allowing the model to capture spatial patterns without requiring complex architectural changes. This resolves the contradiction by improving prediction accuracy through dimensional transformation rather than increasing model complexity.
Solution Approach 2:
The patent changes parameters by transforming feature dimensions and using learned dimension mapping matrices to reparameterize the input space. Instead of modifying the model architecture to be more complex, the approach changes the parameter representation of spatial relationships through dimension conversion, thereby improving prediction accuracy while maintaining relatively simple model structures.
2Ease of operation
If spatial correlation is not considered in the learning model, then the model structure remains simple, but the interpretability deteriorates due to difficulty in identifying effective parameters at specific sites
Solution Approach 1:
By converting features to different dimensions that explicitly represent spatial locations and relationships, the patent makes effective parameters at specific sites more identifiable. The dimension mapping transforms abstract feature spaces into spatially-aware representations, improving interpretability without requiring complex model structures to track spatial dependencies.
3Measurement precision
If dimension mapping is introduced to account for spatial correlation, then the prediction accuracy improves by reflecting actual spatial distributions, but the processing complexity increases
Solution Approach 1:
The patent performs dimension mapping and feature conversion as preliminary actions before the main prediction task. By pre-processing features to encode spatial correlations upfront, the actual prediction model can operate more efficiently with already-structured spatial information, reducing the processing complexity during inference while maintaining high prediction accuracy.
Data Source
AI summary
A non-transitory computer readable recording medium storing a computer program causing a computer to execute a process of acquiring data related to substrate processing, extracting features of acquired data, using a first learning model which has been trained to output features of data in response to an input of the data, converting extracted features into features having a set target dimension, and computing a predicted value by inputting the features with converted dimension to a second learning model, which has been trained to output the predicted value related to the substrate processing in response to an input of the features having the target dimension.