Machine Learning Training Data Selection via Domain Model Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for training machine learning algorithms do not adequately consider the performance of the algorithm with respect to the dimensions and options of the domain model, nor do they address the importance of these elements effectively.
Innovation Solution
A method that involves providing a domain model with domain parameters and values, creating a data model with a training data set, reducing the training data set by removing or modifying data based on domain parameters and values, training a neural network on the reduced data set, and selecting training data for optimal model performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the entire training data set is used for training, then the model performance is improved, but the training time and computational resources increase
Solution Approach 1:
The patent extracts and removes training data points that have low importance or redundancy based on domain model analysis. By identifying and removing unnecessary data points before training, the system reduces the training data set size while maintaining model performance, thus decreasing training time and computational resource requirements.
Solution Approach 2:
The patent applies partial action by selectively including only the most important training data points in the training process. Instead of using the entire data set, the system identifies a subset of critical data points based on domain parameters and their importance scores, achieving effective training with reduced data volume.
2Adaptability or versatility
If dense sampling and synthetic data are used to cover all dimensions and options, then the coverage of domain model is improved, but the data processing complexity and cost increase
Solution Approach 1:
The patent extracts only the essential domain parameters and their important values based on sensitivity analysis and importance scoring. Instead of densely sampling all possible combinations of dimensions and options, the system identifies and extracts the critical subset that truly impacts model performance, reducing data processing complexity while maintaining adequate coverage.
Solution Approach 2:
The patent changes the approach from uniform dense sampling to selective sampling based on parameter importance. By analyzing the importance of different domain parameters and their values, the system adjusts the sampling strategy to focus on critical parameters, thereby reducing the overall data volume and processing complexity while maintaining model coverage.
3Reliability
If all training data are included to ensure comprehensive coverage, then the completeness of training is improved, but the identification of important dimensions and options becomes difficult
Solution Approach 1:
The patent segments the training data set based on domain parameters and their importance levels. By dividing the data into segments based on the significance of different dimensions and options, the system can analyze and identify important parameters more easily while ensuring that critical segments are adequately represented in the training process.
Solution Approach 2:
The patent implements feedback mechanisms through sensitivity analysis and performance evaluation. By systematically removing or hiding specific domain parameters and observing the impact on model performance, the system receives feedback about the importance of different dimensions and options, enabling identification of critical parameters while maintaining training completeness.
Data Source
AI summary
A method for optimized training of a machine learning algorithm. The method includes: providing a domain model that has domain parameters and/or domain values for at least one domain;providing a data model that has a training data set including training data for the at least one domain;removing/hiding/modifying at least one training datum from the training data set depending on at least one domain parameter and/or domain value to provide a reduced training data set;training a neural network based on the reduced training data set to determine a model performance depending on the reduced data set; comparing the determined model performance with a model performance associated with the training data set; selecting training data from the training data set depending on the comparison of the model performances; training the machine learning algorithm based on the selected training data; and providing the trained machine learning algorithm.