Synthetic Data Generation with Anonymity Preservation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Computer-based reasoning systems face challenges in acquiring extensive and specific training data, particularly for rare scenarios, which are costly, difficult, or dangerous to create, and often require anonymization to ensure privacy and regulatory compliance.
Innovation Solution
The techniques involve generating synthetic data using existing training data and target surprisal, where undetermined features are conditioned on previous values or derivatives, and feature bounds are used to ensure the synthetic data differs from the original data, while preserving anonymity and privacy measures like k-anonymity and t-closeness.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If extensive real training data is acquired for rare scenarios, then model training effectiveness is improved, but data acquisition cost and difficulty increase significantly
Solution Approach 1:
The patent generates synthetic training data by copying and transforming existing real data through conditional generation processes. Synthetic data cases are created that mimic the statistical properties and patterns of rare scenarios without requiring actual collection of expensive real-world examples, thus resolving the contradiction between training effectiveness and acquisition difficulty
Solution Approach 2:
The system changes data parameters by generating synthetic samples with modified features while preserving underlying distributions. By adjusting parameters such as feature bounds, conditional probabilities, and data transformations, the system creates diverse synthetic data that maintains training effectiveness while avoiding the costs of real data collection
2Measurement precision
If real training data is used to improve model performance, then training accuracy is improved, but privacy and anonymity compliance deteriorates
Solution Approach 1:
The patent creates synthetic copies of real training data that preserve statistical properties and training value while removing all personally identifiable information. The synthetic data maintains the same distributions, correlations, and patterns needed for accurate model training but cannot be traced back to individual subjects, thus resolving the privacy-accuracy contradiction
Solution Approach 2:
The synthetic data generation system acts as an intermediary between real data and model training. It transforms real data into synthetic representations that serve as a protective layer, allowing models to learn from data that is statistically equivalent to real data but legally and ethically safe to use
3Adaptability or versatility
If diverse training data is generated for rare scenarios, then model robustness is improved, but data generation complexity increases
Solution Approach 1:
The patent segments the data generation process into modular components: conditional probability calculation, feature bound determination, synthetic case generation, and validation steps. This segmentation allows complex diverse data generation to be managed through systematic, reusable modules that handle different scenario types independently
Data Source
AI summary
Techniques for synthetic data generation in computer-based reasoning systems are discussed and include receiving a request for generation of synthetic data based on a set of training data cases. One or more focal training data cases are determined. For undetermined features (either all of them or those that are not subject to conditions), a value for the feature is determined based on the focal cases. In some embodiments, the generated synthetic data may be checked for similarity against the original data, and if similarity conditions are met, it may be modified (e.g., resampled), removed, and/or replaced.