Synthetic Data Generation for Machine Learning Model Expansion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for creating synthetic data are slow and error-prone, particularly when dealing with sensitive data, and they fail to maintain the statistical characteristics of the original data, limiting their utility for training and testing machine learning models.
Innovation Solution
The system generates synthetic data and automatically adds additional classes to create a synthetic classified dataset, which can be used to retrain previously trained machine learning models, thereby allowing for more rapid development and secure training of improved models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If existing methods are used to create synthetic data, then data can be generated for testing and training, but the process is extremely slow and error-prone
Solution Approach 1:
The patent uses Generative Adversarial Networks (GANs) to create synthetic data that copies the statistical characteristics and patterns of real sensitive data without reproducing actual sensitive information. The generator network learns to copy the distribution and relationships in the training data, while the discriminator network verifies the authenticity of generated samples, enabling fast and reliable synthetic data generation
2Object-affected harmful factors
If manual methods are used to desensitize data, then some protection can be achieved, but human error occurs and statistical characteristics are lost
Solution Approach 1:
The patent replaces manual mechanical desensitization processes with an automated neural network-based system. The GAN architecture automatically learns to generate synthetic data that preserves statistical characteristics while ensuring security, eliminating human error and maintaining data fidelity through algorithmic processes rather than manual intervention
3Measurement precision
If production data is used to retrain models with new classes, then model accuracy improves, but sensitive data is exposed to development environment
Solution Approach 1:
The patent introduces synthetic data as an intermediary between the development environment and the sensitive production data. Instead of directly exposing production data to developers, the system generates synthetic training data that mediates the learning process, allowing models to be retrained on new classes while keeping sensitive data isolated in the production environment
4Adaptability or versatility
If new data models are added for new classes, then classification coverage increases, but training time increases considerably
Solution Approach 1:
The patent merges the classification of new data classes into the existing trained model through incremental retraining using synthetic data, rather than training completely separate new models. This combines the knowledge from previously trained classes with new class information in a unified model, reducing training time while maintaining comprehensive classification coverage
Data Source
AI summary
Systems and methods for classifying data are disclosed. For example, a system may include at least one memory storing instructions and at least one processor configured to execute the instructions to perform operations. The operations may include receiving training data comprising a class. The operations may include training a data classification model using the training data to generate a trained data classification model. The operations may include receiving additional data comprising labeled samples of an additional class not contained in the training data. The operations may include creating a synthetic data generator. The operations may include training the synthetic data generator to generate synthetic data corresponding to the additional class. The operations may include generating a synthetic classified dataset comprising the additional class. The operations may include retraining the trained data classification model using the synthetic classified dataset.