Representative Dataset Selection for High-Coverage ML Training
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning (ML) model development is computationally complex and costly due to the use of large initial datasets, and existing methods fail to select representative datasets that accurately reflect the characteristics of the original dataset, leading to models that do not encompass all features of the original dataset.
Innovation Solution
A method for selecting a representative dataset by considering data density and dissimilarity, iteratively adding points with high density and low dissimilarity to a training dataset, adjusting neighborhood size and budget limits to optimize coverage and performance, thereby reducing computational complexity and resource requirements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a large initial dataset is used for ML model training, then model accuracy and coverage are improved, but computational complexity and resource requirements increase
Solution Approach 1:
The patent extracts a representative subset of data points from the large initial dataset based on density and dissimilarity metrics. This subset contains only the most informative samples needed for model training, eliminating redundant data while preserving model accuracy and reducing computational burden.
Solution Approach 2:
The patent changes the parameter of dataset size by selecting a smaller representative subset instead of using the full large dataset. This parameter change is guided by density and dissimilarity calculations that ensure the subset maintains the essential characteristics of the original data distribution.
2Device complexity
If a smaller representative dataset is selected to reduce computational complexity, then resource usage decreases, but model coverage and accuracy deteriorate
Solution Approach 1:
The patent employs feedback mechanisms by iteratively training models on selected subsets and evaluating their performance. The selection process uses density and dissimilarity metrics that are continuously refined based on model performance feedback, ensuring the subset maintains high coverage and accuracy.
Solution Approach 2:
The patent replaces the mechanical approach of using all available data with a systematic selection process based on density and dissimilarity calculations. This substitution uses computational metrics to identify and select only the most valuable data points, achieving high model coverage with reduced dataset size.
3Device complexity
If existing data selection methods are used to reduce dataset size, then computational resources are saved, but the selected dataset fails to accurately reflect original data characteristics
Solution Approach 1:
The patent performs preliminary actions by calculating density and dissimilarity metrics for all data points before selection. This pre-processing step identifies the most representative samples in advance, ensuring the selected subset accurately reflects the original data characteristics before the actual model training begins.
Solution Approach 2:
The patent changes the selection criteria parameter from random or simple sampling to density-based and dissimilarity-based selection. This parameter change ensures that the selected subset maintains the essential statistical properties and distribution characteristics of the original large dataset.
Data Source
AI summary
Providing a representative dataset from an initial dataset by accessing a dataset associated with a machine learning model, receiving input parameters associated with the representative dataset selection, the input parameters including an evaluation metric, determining a density of a plurality of datapoints associated with the dataset, training a first iteration of a machine learning model using a first data point selected according to the density, determining a first value of the evaluation metric for the first iteration of the machine learning model, generating a representative subset based on the first value of the evaluation metric value, and providing the representative dataset and a final machine learning model trained using the representative dataset.