Self-Learned Model Selection Reference for Faster AutoAI
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current automated machine learning (AutoML) model generation and selection implementations often impose limitations by failing to leverage user data structures across scenarios and neglect contextual relevance, leading to inefficient and slow model building processes.
Innovation Solution
Employ self-learned prior knowledge through descriptive and multinomial logistic regression models to build Model Selection Acceleration Mechanism (MSAM) models, which predict top algorithms for given data sets by detecting similar patterns and avoiding duplicate workloads.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional AutoML model selection processes are used, then model selection can be performed, but the process is inefficient and fails to leverage prior knowledge from similar data patterns
Solution Approach 1:
The system performs preliminary actions by identifying and storing data pattern groups and their associated successful model configurations in advance. When a new data set arrives, the system first checks if similar patterns exist in the stored knowledge base, allowing it to leverage prior results without re-evaluating all models from scratch, thus significantly reducing model selection time
Solution Approach 2:
The system creates copies of successful model configurations and hyperparameter settings from previously analyzed data patterns. Instead of training and evaluating models independently for each new data set, the system copies proven configurations from similar patterns and adapts them, dramatically improving productivity while reducing the time needed for model selection
2Productivity
If independent model evaluation is performed for each data set, then model selection can be done, but duplicate workloads are created across similar data patterns
Solution Approach 1:
The system merges the model evaluation process across multiple data sets by identifying shared data patterns. When similar patterns are detected, the system combines their evaluation results and leverages the knowledge base to avoid redundant model training and evaluation, reducing duplicate workloads and optimizing the use of computational resources
Solution Approach 2:
The system implements self-service by automatically identifying data patterns, comparing them against the knowledge base, and selecting appropriate model configurations without requiring manual intervention or repeated full evaluations. This automated self-service approach reduces both time loss and computational energy waste
Data Source
AI summary
A method, system, and computer program product for self-learning reference mechanisms for model selection in AutoAI. The method identifies a set of data summary statistics within a data set. A data pattern group is identified within the set of data summary statistics. The data pattern group is determined to be mature. A model selection acceleration mechanism (MSAM) model is generated based on the data pattern group. The method predicts a set of top-k models for the data set based on the MSAM model.