Multi-layered Ensemble Learning for Automated Data Labeling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing artificial intelligence systems face challenges in efficiently updating models deployed in commercial use, as real-use data often lacks labels, requiring time, money, and specialized expertise to manually label, which delays model retraining and deployment.
Innovation Solution
A multi-layered machine learning system is introduced, utilizing multiple base models and a fusion model to automate the labeling of unlabeled data, where base models are specialized differently to cover complementary aspects of the problem space and the fusion model combines intermediate outputs to generate accurate final labels.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual labeling of real-use data is performed, then labeling accuracy is improved, but time consumption and cost increase
Solution Approach 1:
The system enables automated self-labeling of real-use data by training base models on initially labeled data, then using these models to generate labels for unlabeled data. This self-service mechanism eliminates the need for continuous manual labeling while maintaining labeling quality through the ensemble fusion process.
Solution Approach 2:
Base models serve as intermediaries between the initial labeled data and the final labeling results. These models process unlabeled data and generate intermediate labels that are then fused by the fusion model to produce final labels, reducing direct human intervention while preserving accuracy.
2Measurement precision
If manual labeling by specialists is performed, then labeling quality is improved, but cost increases
Solution Approach 1:
The system replaces expensive specialist labeling with automated self-labeling using trained base models. The models independently generate labels for real-use data, eliminating the need for specialized human expertise while maintaining labeling quality through the ensemble approach.
Solution Approach 2:
The system creates copies of labeling capability through multiple base models trained on initial labeled data. These model copies can independently label data without requiring original specialists, significantly reducing cost while maintaining quality through fusion of multiple model outputs.
3Loss of energy
If model retraining is delayed, then resource consumption is reduced, but model performance deteriorates
Solution Approach 1:
The system enables continuous automated retraining by generating labels for real-use data without manual intervention. Base models continuously process new data, generate labels, and update the ensemble, ensuring model performance stays aligned with actual usage patterns without requiring scheduled resource-intensive retraining campaigns.
Solution Approach 2:
The system performs preliminary labeling of real-use data using base models before full model retraining. This preliminary action prepares labeled data in advance, enabling faster and more efficient retraining cycles while maintaining model performance alignment with current usage patterns.
4Measurement precision
If multiple base models are used, then labeling accuracy is improved, but system complexity increases
Solution Approach 1:
The fusion model merges outputs from multiple specialized base models into a unified labeling result. By combining the strengths of different base models through fusion techniques (such as weighted averaging or voting), the system achieves higher labeling accuracy while managing complexity through a single fusion layer that coordinates the multiple models.
Data Source
AI summary
A method includes providing input data to a plurality of base models to generate a plurality of intermediate outputs. The base models are non-linear in that different base models are specialized differently such that the different base models are complementary to one another. Each of the base models is generated using a different base classification algorithm in a multi-layered machine learning system. The method also includes processing the intermediate outputs using a fusion model to generate a final output associated with the input data. The fusion model is generated using a meta classification algorithm in the multi-layered machine learning system. The method may also include training the classification algorithms, where training data used by each of at least one of the base classification algorithms is selected based on an uncertainty associated with at least one other of the base classification algorithms.