Machine Learning Data Generation with Human Execution Feedback
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data generation methods for machine learning do not account for the evaluation of learning data accuracy, which can impact the classification process, leading to suboptimal results.
Innovation Solution
A data generation method that acquires both result and execution information from human classification tasks, generating learning data with evaluation information to improve the accuracy of classification models by selectively using highly evaluated data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If learning data is generated without evaluation information, then the data generation process is simple, but the classification accuracy of the learned model deteriorates
Solution Approach 1:
The patent implements feedback by acquiring execution information (time, condition, subjective opinion) from the living being's classification process and using it to generate evaluation information. This feedback loop allows the system to assess the quality of learning data based on actual execution characteristics, thereby improving classification accuracy while maintaining a manageable data generation process through automated evaluation metrics.
Solution Approach 2:
The patent introduces evaluation information as an intermediary element that bridges the gap between raw execution information and learning data quality assessment. This intermediary layer processes execution information (time, condition, subjective opinion) into structured evaluation metrics, enabling systematic quality control without requiring complex direct analysis of all execution parameters.
2Measurement precision
If all learning data is used without evaluation, then the data processing is efficient, but the quality of classification results deteriorates
Solution Approach 1:
The patent applies local quality by evaluating and selecting learning data based on specific execution characteristics (time, condition, subjective opinion) rather than treating all data uniformly. High-quality data points meeting certain evaluation criteria are selected for training, while lower quality data is excluded, thereby improving classification result quality without requiring processing of all available data.
Solution Approach 2:
The patent changes the parameter set used for data selection by introducing evaluation information derived from execution parameters (time, condition, subjective opinion). This parameter transformation enables the system to identify and prioritize high-quality learning data, improving classification results while maintaining processing efficiency through targeted data selection rather than comprehensive data processing.
3Measurement precision
If execution information is collected in detail, then the evaluation accuracy of learning data improves, but the information collection burden increases
Solution Approach 1:
The patent applies universality by using a single information collection mechanism that simultaneously captures multiple types of execution information (time, condition, subjective opinion) from the living being's classification process. This multi-functional approach enables comprehensive evaluation accuracy improvement without requiring separate specialized systems for each type of information, thereby managing system complexity effectively.
Data Source
AI summary
A data generation method includes a first acquisition step, a second acquisition step, and a generation step. The first acquisition step includes acquiring result information about a result of a classification executed by a living being on a target. The second acquisition step includes acquiring execution information about execution of the classification. The generation step includes generating data for machine learning based on the result information and the execution information. The data for machine learning includes learning data and evaluation information about evaluation of the learning data.