Multi-label Ensemble Learning via Multi-center Prediction Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing ensemble techniques for predicting future health states in medical contexts face ambiguities in determining correct answer labels and weighting predictors, leading to potential noise learning and reduced accuracy due to unclear criteria for selecting and weighting prediction values.
Innovation Solution
A method and apparatus for learning a multi-label ensemble based on multi-center prediction accuracy, which involves collecting prediction values, calculating prediction errors, generating error-based weight scores, optionally selecting and normalizing these scores to create weight labels, and learning an ensemble weight prediction model to assign weights to predictors based on their accuracy, thereby clarifying the correct answer and reducing ambiguity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If ensemble techniques are used to synthesize prediction results from multiple predictors, then prediction accuracy is improved by removing noises through arithmetic mean or weighted sum, but ambiguity arises in determining correct answer labels and weighting predictors
Solution Approach 1:
The patent implements feedback by using the ensemble model's own prediction results to generate soft labels, which are then used to retrain and refine the ensemble. This closed-loop feedback mechanism resolves ambiguity by allowing the system to learn from its own predictions and continuously improve the correctness of labels without requiring external ground truth
Solution Approach 2:
The ensemble model serves itself by generating its own training labels through its prediction outputs. The soft labels produced by the ensemble are used as targets for retraining, enabling the system to self-improve without external intervention. This self-service approach eliminates the need for ambiguous external label sources
2Reliability
If multiple prediction models are synthesized using arithmetic mean or weighted sum, then variability in prediction is reduced, but clarity in weighting criteria is lost
Solution Approach 1:
The system automatically determines optimal weights through self-training using soft labels generated from ensemble predictions. This self-service mechanism eliminates the need for manual weight assignment or complex external weighting criteria, allowing the model to autonomously learn the most appropriate weights for each predictor based on their performance
Solution Approach 2:
The patent performs preliminary training of individual prediction models before ensemble synthesis. This preliminary action ensures that each model is properly calibrated and ready for ensemble integration, reducing the need for complex post-hoc weighting adjustments and simplifying the overall weighting process
Data Source
AI summary
Disclosed herein a method and apparatus for learning a multi-label ensemble based on multi-center prediction accuracy. According to an embodiment of the present disclosure, there is provided a multi-label ensemble learning method comprising: collecting a prediction value for learning data for each of a plurality of prediction models; calculating a prediction error of each of the prediction models using the prediction value of each of the prediction models and a correct answer prediction value; generating a weight label for each of the prediction models based on the prediction error; and learning an ensemble weight prediction model for predicting a weight of each of the prediction models using the weight label.