Classifier-Guided RNN Synthetic Data Generation with Token Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Machine learning systems face challenges in generating highly structured synthetic data, such as log files and database records, due to occasional small mistakes that render the data invalid or change its semantics, requiring additional post-processing and validation.
Innovation Solution
A recurrent neural network (RNN) is used to generate synthetic data, combined with a classifier to automatically identify and correct errors by replacing sequences matching predefined strings with tokens, and updating the RNN's vocabulary to include these tokens.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a recurrent neural network is used to generate highly structured synthetic data, then the data can be generated automatically, but small mistakes occur that render the data invalid or change its semantics
Solution Approach 1:
The system implements feedback by using a classifier to evaluate generated synthetic data and identify errors. The classifier provides feedback signals that guide the RNN to correct mistakes in generating highly structured data, thereby improving data validity while maintaining automatic generation capability
Solution Approach 2:
The system performs preliminary action by using the classifier to predict and identify potential errors before they propagate through the data generation process. This allows the RNN to correct mistakes proactively during generation rather than requiring post-processing validation
2Measurement precision
If post-processing and validation are added to identify and remove invalid data, then data accuracy improves, but resources and time required to generate synthetic data increase
Solution Approach 1:
The classifier performs preliminary error identification during the data generation process itself, rather than requiring separate post-processing validation steps. This integrates accuracy checking into the generation workflow, improving data accuracy without proportionally increasing total generation time
Solution Approach 2:
The system implements self-service by enabling the RNN to automatically correct its own errors through the classifier's feedback mechanism. The generated data undergoes self-validation and self-correction without requiring external post-processing intervention, thereby improving accuracy while minimizing additional time costs
Data Source
AI summary
The present disclosure provides systems and methods for synthetic data generation. A recurrent neural network can be trained for synthetic data generation by obtaining a sequence of elements and determining, using a classifier, that the sequence corresponds to a token. In response to the determination, a recurrent neural network configured to use a first vocabulary including the elements can be modified to use a second vocabulary, the second vocabulary including the token and the first vocabulary. The modified recurrent neural network can be trained using the token and the sequence of elements. The trained recurrent neural network can be used to generate synthetic data. A classifier can detect sequences of elements in the synthetic data corresponding to tokens. The tokens can replace the sequences of elements in the generated synthetic data and can be provided to the trained recurrent neural network to continue synthetic data generation.