AI Training Data Sampling and Hyperparameter Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Machine learning models trained with incomplete or poor-quality data sets produce inaccurate results, and there is a need for techniques to improve the accuracy of these models, especially when important ranges of user preferences are poorly represented, and to enhance the relevance of input variable data sets in large and diverse data environments.
Innovation Solution
A computer-implemented method that involves loading a training data set with multiple bins, applying under-sampling and over-sampling techniques, merging the updated bins, and training the machine learning model, while also generating and filtering input variables to emphasize their frequency in data objects and de-emphasize their frequency in groups of objects, and automatically determining optimal hyperparameters for the model.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional machine learning training is used with incomplete or poor-quality data sets, then training process is simple, but model accuracy deteriorates
Solution Approach 1:
The patent applies preliminary action by performing data sampling adjustments (under-sampling and over-sampling) before model training. The system identifies important ranges in the training data set and adjusts the distribution of data points in these ranges beforehand, so that when training proceeds, the model receives balanced quality data without requiring complex real-time adjustments during training.
Solution Approach 2:
The patent changes parameters of the training data set by modifying the distribution of data points across different bins. It adjusts sampling rates for different ranges of input variables, transforming the data distribution parameters to emphasize important ranges while reducing emphasis on less important ranges, thereby improving model accuracy without fundamentally changing the training architecture.
2Reliability
If automated techniques are applied to improve input variable relevance, then data quality improves, but processing complexity increases
Solution Approach 1:
The patent segments the input variable data into different bins based on frequency analysis. It divides the data range into multiple segments (bins) and applies different sampling strategies to each bin. This segmentation allows the system to manage complex data processing by breaking it into manageable parts, where each bin can be processed independently with appropriate sampling adjustments.
Solution Approach 2:
The patent replaces manual or mechanical data processing approaches with automated frequency-based filtering. Instead of relying on manual assessment of input variable relevance, the system automatically analyzes frequency patterns in the data and adjusts sampling accordingly, substituting automated computational analysis for manual data evaluation processes.
3Productivity
If frequency filtering is applied to emphasize important scalar values, then model performance improves, but computational time increases
Solution Approach 1:
The patent applies partial action by performing frequency filtering and sampling adjustment on only the most critical portions of the data. It identifies important ranges where frequency filtering has the greatest impact on model performance and concentrates computational effort there, rather than uniformly processing all data. This allows the system to achieve significant performance improvements with reduced overall computational time.
Data Source
AI summary
A non-transitory computer-readable medium includes executable instructions including loading a training data set including a first bin and a second bin, applying an under-sampling technique to elements of the first bin to generate an updated first bin, applying an over-sampling technique to elements of the second bin to generate an updated second bin, generating an updated training data set by merging the updated first bin and the updated second bin, loading baseline hyperparameters, configuring a machine learning model with the baseline hyperparameters, providing the updated training data set to the machine learning model to determine baseline performance metrics, determining whether the baseline performance metrics are above a threshold, saving the baseline hyperparameters as optimal hyperparameters in response to determining that the baseline performance metrics are above the threshold, configuring the machine learning model with optimal hyperparameters, and providing input variables to the machine learning model to generate output variables.