Meta-Learning Model Selection via Configuration Space Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current machine learning techniques require extensive configuration tuning and resource-intensive processes to achieve optimal performance, making them inefficient and resource-heavy.
Innovation Solution
A method and apparatus that utilize meta-learning to optimize the configuration space by extracting meta-features from datasets, calculating performance based on various configurations, and executing meta-learning to refine the configuration space, thereby reducing the search space and resource requirements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If extensive configuration tuning is performed to achieve optimal machine learning performance, then model accuracy is improved, but computing time and resource consumption increase
Solution Approach 1:
The patent applies preliminary action by pre-calculating and storing performance metrics for various machine learning configurations in a database before actual model selection. When a new dataset arrives, the system retrieves pre-computed performance data and matches it with dataset characteristics, avoiding the need to test every configuration from scratch. This significantly reduces computing time while maintaining the ability to identify optimal configurations.
Solution Approach 2:
The patent uses copying by creating a configuration space model that represents the performance characteristics of multiple machine learning configurations without physically executing each one. The system copies performance patterns from previously analyzed configurations and uses them to predict optimal settings for new datasets, reducing the need for exhaustive computational testing.
2Measurement precision
If a comprehensive configuration space is searched to find optimal machine learning models, then model performance is improved, but resource consumption increases
Solution Approach 1:
The patent extracts only the essential performance characteristics and configuration parameters needed for model selection, storing them in a structured configuration space database. Instead of processing entire datasets and models, the system extracts key features and performance metrics, then uses these extracted elements to determine optimal configurations, significantly reducing computational resources while maintaining selection accuracy.
Solution Approach 2:
The system performs preliminary analysis to build a configuration space database that pre-contains performance information for various machine learning configurations. This preliminary work allows the system to quickly query and compare configurations without re-running expensive computational experiments, reducing resource consumption during actual model selection tasks.
3Measurement precision
If multiple configurations are evaluated to select the best machine learning model, then selection accuracy is improved, but processing time increases
Solution Approach 1:
The patent performs preliminary evaluation of multiple configurations and stores their performance characteristics in advance. When a selection is needed, the system retrieves pre-computed performance data and compares it with current dataset characteristics, maintaining high selection accuracy while dramatically reducing the time required for the actual selection process.
Solution Approach 2:
The system creates a configuration space representation that copies the essential performance patterns of multiple configurations without requiring their full execution. This copied representation enables rapid comparison and selection while preserving the accuracy benefits of evaluating multiple configurations.
Data Source
AI summary
A method of selecting a model of machine learning executed by a processor is provided. The method includes: receiving at least one data-set; configuring a configuration space for machine learning of the at least one data-set; extracting, from the at least one data-set, a meta-feature including quantitative information about the data-set; calculating performance of the machine learning for the at least one data-set based on a plurality of configurations included in the configuration space; executing meta-learning based on the meta-feature, the plurality of configurations, and the calculated performance; and optimizing the configuration space based on a result of executing the meta-learning.