Integrated Model Generation Using Iterative Candidate Regression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing meta-learning methods for generating integrated models from multiple trained models are time-consuming due to the consideration of all allowable combinations, leading to increased generation time as the number of trained models grows.
Innovation Solution
An information processing apparatus and method that includes a selector, training unit, evaluation metric calculator, and regression expression generator to iteratively select, train, and evaluate candidate models until a predetermined termination condition is satisfied, generating a regression expression with higher prediction accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If all allowable combinations of plural trained models are considered to generate an integrated model, then the prediction accuracy is improved, but the time required to generate the integrated model increases as the number of trained models increases
Solution Approach 1:
The patent segments the exhaustive combination search into multiple iterations. In each iteration, only a subset of candidate models is selected and evaluated, rather than considering all possible combinations at once. This divides the large search space into manageable segments that can be processed sequentially, reducing the time burden while still exploring diverse model combinations.
Solution Approach 2:
The patent applies partial action by evaluating only a portion of all possible model combinations in each iteration. Instead of exhaustively testing every combination, the system selectively evaluates promising candidates based on preliminary criteria, achieving satisfactory prediction accuracy without the computational overhead of complete enumeration.
2Measurement precision
If the number of trained models increases to improve prediction accuracy, then the quality of the integrated model is improved, but the complexity of the generation process increases
Solution Approach 1:
The patent introduces dynamic adaptation in the model selection process. The selection criteria and candidate pool are adjusted based on performance feedback from previous iterations. This dynamic approach allows the system to handle increasing numbers of trained models efficiently by adapting the search strategy rather than using a fixed complex procedure.
Solution Approach 2:
The system implements feedback mechanisms where the performance of integrated models from previous iterations informs the selection of candidates in subsequent iterations. This feedback loop enables the system to learn from past results and refine its search strategy, managing the complexity of handling multiple trained models through iterative improvement rather than static complex processing.
Data Source
AI summary
An apparatus has a first function of selecting, from individual prediction models, candidate models, a second function of training the candidate models using a training dataset, a third function of applying a test dataset to a candidate integrated model to calculate an evaluation metric, and a fourth function of generating a regression expression between the evaluation metric and the candidate combination. A controller repeats a sequence of the first to fourth functions while the first function selects new individual models for each sequence until a termination condition is satisfied. The regression expression generated with an additional value of the evaluation metric from the new candidate combination for each sequence has a higher prediction accuracy on average as the number of the repeated sequences increases. A determiner determines an integrated model based on the candidate integrated models and values of the evaluation metric after the termination condition is satisfied.