Weighted Data Augmentation for Machine Learning Model Training
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Machine learning models often suffer from overfitting or underfitting due to inadequate quality and quantity of training data, which negatively impacts their performance.
Innovation Solution
A computer-implemented method that applies predetermined augmentations to a sample set of training data to create an augmented sample set, trains a model with the augmented data, determines the model's performance, and assigns weights to the augmentations based on performance, deciding whether to apply these augmentations to the larger training data set.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If data augmentation is applied to increase training data quantity, then model performance is improved, but computational resources and training time are increased
Solution Approach 1:
The system performs preliminary evaluation of data augmentations on a sample set before applying them to the full training data set. By testing augmentations on a smaller subset first and evaluating their impact on model performance, the system identifies effective augmentations in advance, avoiding the need to test all augmentations on the complete data set and thus reducing overall training time while maintaining performance improvements.
2Reliability
If multiple data augmentations are applied to improve model accuracy, then overfitting is reduced, but system complexity increases
Solution Approach 1:
The system applies data augmentations selectively rather than all at once. It evaluates augmentations individually or in small groups on a sample set, identifies the most effective ones, and applies only those to the full training data set. This partial application approach reduces system complexity by avoiding the need to manage and evaluate all possible augmentations simultaneously, while still achieving the benefit of reduced overfitting through the application of effective augmentations.
3Reliability
If data augmentation is applied to enlarge training data set, then model generalization is improved, but computational resources are increased
Solution Approach 1:
The system segments the training process into two phases: (1) evaluation phase on a small sample set where multiple augmentations are tested, and (2) application phase on the full training data set where only the most effective augmentations are applied. This segmentation allows the system to identify high-value augmentations using minimal computational resources in the evaluation phase, then apply only those selected augmentations in the application phase, thereby improving model generalization while controlling computational resource usage.
Data Source
AI summary
A computer-implemented method according to one embodiment includes applying a predetermined augmentation to the sample set of training data to create an augmented sample set, training a model with the augmented sample set, determining a performance of the trained model, and assigning a weight to the predetermined augmentation for the training data set based on the determined performance. A determination is made as to whether to apply the predetermined augmentation to a larger training data set before the training data set is applied to the model, based on the weight assigned to the predetermined augmentation.