Synthetic Data Generation via Feature Importance Queries
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data-driven approaches for synthetic data generation in machine learning face scalability issues due to exponential computational overhead when dealing with high-dimensional data, especially when k-way marginal queries exceed 2, which complicates privacy preservation and data utility.
Innovation Solution
A method that prioritizes features with high importance scores to generate marginal queries, reducing the number of queries and noise required for privacy preservation, allowing for computationally practicable synthetic data generation that maintains data utility and privacy constraints.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If k-way marginal queries are used to generate synthetic data with higher statistical accuracy, then the accuracy of synthetic data improves, but the computational overhead grows exponentially
Solution Approach 1:
The patent extracts only the most important features based on their contribution to the machine learning model output, and generates marginal queries only for these selected features. This extraction approach reduces the number of features k from the total number of features, thereby reducing the exponential computational overhead while maintaining statistical accuracy through focus on high-impact features
Solution Approach 2:
The patent applies local quality by differentiating features based on their importance scores and selectively processing only those features that contribute most significantly to the model output. This creates a non-uniform approach where high-importance features receive full processing while low-importance features are excluded, optimizing the balance between accuracy and computational efficiency
2Measurement precision
If k-way marginal queries are used to generate synthetic data with higher statistical accuracy, then the accuracy of synthetic data improves, but the amount of noise required for privacy preservation increases significantly
Solution Approach 1:
The patent extracts only the most important features based on their contribution to the machine learning model output, and generates marginal queries only for these selected features. This extraction approach reduces the number of features k from the total number of features, thereby reducing the exponential computational overhead while maintaining statistical accuracy through focus on high-impact features
Solution Approach 2:
The patent applies local quality by differentiating features based on their importance scores and selectively processing only those features that contribute most significantly to the model output. This creates a non-uniform approach where high-importance features receive full processing while low-importance features are excluded, optimizing the balance between accuracy and computational efficiency
3Adaptability or versatility
If all features are included in marginal queries to maintain data utility, then the utility of synthetic data improves, but the computational complexity becomes impractical for high-dimensional data
Solution Approach 1:
The patent changes the parameter of feature selection by using importance scores to rank and select features. Instead of including all features, the system selects a subset based on their importance scores, transforming the problem from processing all features to processing only the most critical ones, thus reducing computational complexity while maintaining data utility
Solution Approach 2:
The patent applies partial action by processing only a subset of features (those with highest importance scores) rather than all features. This partial processing approach is sufficient to maintain data utility for machine learning training while dramatically reducing computational complexity for high-dimensional data
Data Source
AI summary
A computer-implemented method including: receiving a trained machine learning model; extracting a set of features associated with the machine learning model, wherein each of the extracted features is assigned a feature importance score which represents a relative explanatory power of the feature with respect to an output of the machine learning model; generating a set of marginal queries based, at least in part, on a selected subset of the features having a highest the feature importance score; performing a measurement of the set of marginal queries on a source database, to obtain measurements of the set of marginal queries on the source database; and using the measurements to generate synthetic data that matches the measurements.