Machine Learning Model Training Data Selection and Noise Injection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning methods, such as stacking, require repeated processing of multiple machine learning models to generate training data for a second model, making the process inefficient and data-intensive.
Innovation Solution
A method that selects stable data cases based on appearance frequency to generate a first machine learning model, and then uses the model's output to create a training data set for a second model, reducing the need for multiple model creations and improving data efficiency by adding noise to unstable cases.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If stacking method is used with multiple machine learning models to improve inference accuracy, then the accuracy improves, but the processing time and computational complexity increase
Solution Approach 1:
The patent applies preliminary action by pre-processing training data to identify and remove unstable cases before model training. This preparation work is done in advance to reduce the computational burden during the stacking process, allowing multiple models to be trained more efficiently without sacrificing the accuracy improvements that stacking provides
Solution Approach 2:
The patent extracts and removes unstable training data cases from the dataset before training machine learning models. By taking out these problematic cases that would require extensive processing and correction, the system reduces the time and computational resources needed while maintaining the accuracy benefits of the stacking approach
2Measurement precision
If repeated processing of multiple machine learning models is performed to generate training data, then the model accuracy improves, but the data processing efficiency decreases
Solution Approach 1:
The system performs preliminary identification and removal of unstable training data cases before the repeated model training process. This advance preparation reduces the amount of data that needs to be processed through multiple modeling iterations, improving data processing efficiency while preserving the accuracy improvements gained from stacking
Solution Approach 2:
The patent extracts unstable cases from the training dataset before initiating the repeated processing of multiple machine learning models. By removing these problematic cases in advance, the system reduces the computational workload and improves processing efficiency while maintaining the ability to achieve high model accuracy through stacking
3Reliability
If training data is divided into multiple subsets for stacking, then the model robustness improves, but the complexity of the training process increases
Solution Approach 1:
The patent extracts and removes unstable training data cases before dividing the remaining data into subsets for stacking. This preliminary extraction simplifies the training process by eliminating cases that would require complex error correction mechanisms, while the subset division maintains model robustness through ensemble learning
Solution Approach 2:
The system performs preliminary cleaning of the training dataset by removing unstable cases before the stacking process begins. This advance preparation reduces the complexity of the training process by eliminating data that would require sophisticated error handling, while the subsequent subset division maintains model robustness
Data Source
AI summary
A non-transitory computer-readable storage medium storing machine learning program that causes a computer to execute a process, the process includes selecting a plurality of data from a first training data group based on an appearance frequency of first data attached with a first label, the first data being included in the first training data group; generating a first machine learning model by training by the plurality of data; and generating a second training data group obtained by combining the first training data group and an output by the first machine learning model when the first data is input.