ML Model Training for Synthetic Data Generation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for generating synthetic data lack efficiency and flexibility, as they often require significant storage and cannot be easily scaled to meet varying data needs.
Innovation Solution
A method for training machine learning models to generate synthetic data, which involves providing an original dataset, determining objective and evaluation functions, and using these functions to optimize the models for generating synthetic data of desired size and quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If traditional synthetic data generation methods are used, then data generation is possible, but storage requirements are significant and scalability is limited
Solution Approach 1:
The patent creates synthetic data copies that replicate the statistical properties and patterns of original data without storing the original large datasets. The synthetic data serves as a lightweight replica that can be generated on-demand, reducing storage requirements while maintaining data utility for training and analysis purposes
Solution Approach 2:
The system implements dynamic synthetic data generation where data is created on-demand based on specified parameters and requirements. Rather than pre-generating and storing fixed datasets, the system can dynamically produce synthetic data of desired size and characteristics, improving scalability and storage efficiency
2Adaptability or versatility
If traditional synthetic data generation methods are used, then data can be generated, but flexibility and scalability to meet varying data needs are poor
Solution Approach 1:
The patent develops a universal synthetic data generation system that can handle multiple data types, formats, and requirements through a single platform. The system accepts various input parameters and configurations to generate different kinds of synthetic data, making it adaptable to diverse needs without requiring separate specialized systems for each data type
Solution Approach 2:
The system enables flexible control over synthetic data generation by allowing users to modify parameters such as data size, statistical properties, distribution characteristics, and domain-specific constraints. By changing these parameters, the system can adapt to varying data needs without fundamental system changes, enhancing versatility while maintaining manageable complexity
3Measurement precision
If ML models are trained to generate synthetic data, then systematic and accurate generation is achieved, but training complexity and computational resources increase
Solution Approach 1:
The patent implements feedback mechanisms during the ML model training process where the generated synthetic data is evaluated against the original data's statistical properties and patterns. This feedback loop allows the model to iteratively improve its accuracy in replicating data characteristics, achieving high measurement precision while managing training complexity through guided optimization
Solution Approach 2:
The system performs preliminary analysis and preparation of the original data before training the ML model, including statistical characterization, pattern identification, and parameter extraction. This preliminary action provides the model with structured guidance during training, reducing the complexity of the learning process while ensuring accurate replication of data properties
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Disclosed is a method for training a set of one or more machine learning (ML) models for generation synthetic data, the method comprising: providing a dataset, referred to as original dataset, comprising records, wherein each record of the records comprises values of the set of attributes; determining at least one objective function and an evaluation function for training the set of one or more ML models; using the original dataset for training each ML model of the set of ML models for optimizing a respective objective function of the at least one objective function, wherein the evaluation function is used for evaluating the ML model during training the ML model or after training the set of one or more ML models; providing at least one of the set of trained ML models.