Machine Learning Model Training Data Filtering for Reaction Speed
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for generating machine learning models for controlling mobile objects, such as vehicles and drones, often fail to produce models capable of implementing control at appropriate reaction speeds due to varying driver capabilities and reaction times in training data, leading to dispersed reaction speeds in learning data.
Innovation Solution
A method that prioritizes data sets in machine learning where the reaction speed to control commands matches a predetermined condition, ensuring the model is trained on data with more appropriate reaction speeds, using a neural network-based control model to derive control commands that match correct answer data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If machine learning is performed using all available training data sets including those with inappropriate reaction speeds, then the quantity of training data is increased, but the reaction speed of the trained model becomes dispersed and inappropriate
Solution Approach 1:
The patent applies parameter changes by filtering training data sets based on reaction speed parameters. Specifically, it calculates reaction speeds for each data set and selectively uses only those data sets whose reaction speeds fall within a predetermined appropriate range, thereby changing the parameter composition of the training data to achieve both sufficient quantity and high reliability
Solution Approach 2:
The patent applies local quality by differentiating between different quality levels of training data. Instead of treating all training data uniformly, it identifies and selects only the high-quality data sets (those with appropriate reaction speeds) for training, giving them priority while excluding or reducing the weight of low-quality data sets with inappropriate reaction speeds
2Reliability
If data sets with inappropriate reaction speeds are excluded from training, then the reaction speed appropriateness is improved, but the quantity of available training data decreases
Solution Approach 1:
The patent changes the parameter selection criteria by establishing a predetermined reaction speed range as a filter condition. This parameter-based filtering ensures that only data sets meeting the reaction speed requirement are used, maintaining high reliability while preserving all data sets that satisfy the condition
Solution Approach 2:
The patent applies partial action by selectively using only the necessary portion of available data sets - specifically those with appropriate reaction speeds - rather than excluding all potentially useful data. This approach ensures sufficient training data quantity by including all data sets that meet the reaction speed criterion
Data Source
AI summary
A model generation method according to one aspect of the present disclosure includes a computer acquiring a plurality of data sets each including a combination of training data and correct answer data and performing machine learning of a control model using the acquired plurality of data sets. Using the plurality of data sets includes preferentially using data sets for which reaction speed with respect to an event of control commands indicated by the correct answer data is evaluated as more appropriate as a result of the reaction speed matching a predetermined condition.