Counter Data Generation for Balanced Dataset Creation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Deep learning models trained with single-class datasets lack sufficient information to accurately classify data points, leading to inaccurate predictions and undermining their effectiveness in binary classification tasks, as they fail to learn from differences and mistakes due to the absence of counter-data.
Innovation Solution
A method to generate a dual-class dataset from a single-class dataset by using unlabeled context data points, which are labeled and iteratively refined to create a balanced dataset, allowing the model to learn from both classes and improve prediction accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a single-class dataset is used for training, then the training process is simpler and faster, but the model's prediction accuracy deteriorates due to lack of counter-data
Solution Approach 1:
The patent applies preliminary action by pre-processing unlabeled context data points before training the model. The context data points are collected and prepared in advance, then used to generate counter-data samples that are integrated into the training dataset. This preliminary preparation enables the model to learn from both positive and negative examples, improving prediction accuracy without significantly increasing training time.
Solution Approach 2:
The patent uses unlabeled context data points as an intermediary to bridge the gap between single-class training data and dual-class learning requirements. These context data points serve as a mediator that is processed to generate counter-data samples, allowing the model to indirectly learn from the absent class without requiring direct access to labeled negative examples during data collection.
2Measurement precision
If counter-data is added to create a dual-class dataset, then the model's ability to distinguish between classes improves, but the dataset complexity and processing time increase
Solution Approach 1:
The patent applies self-service by using the model's own predictions on unlabeled context data to generate counter-data samples. The model processes context data points, and based on prediction confidence thresholds, automatically identifies and labels samples as counter-data. This self-service approach eliminates the need for manual data annotation or complex external data collection processes, reducing dataset processing complexity while maintaining classification accuracy.
3Measurement precision
If iterative refinement of context data is performed, then the quality of the dual-class dataset improves, but the training time and computational resources increase
Solution Approach 1:
The patent applies partial action by performing iterative refinement only on a subset of context data points rather than processing the entire dataset repeatedly. The refinement process uses confidence thresholds to identify only those samples that benefit from re-processing, applying partial updates to the dataset. This approach improves dataset quality through targeted refinement while minimizing the time and computational resources required compared to exhaustive iterative processing.
Data Source
AI summary
A method for generating a dual-class dataset is disclosed. A single-class dataset and a context dataset are obtained. The context dataset can be labeled. A model can be trained using the combination of the single-class dataset and the labeled context dataset. The model can be run on the context dataset. The data points that are classified the same as the data points included in the single-class dataset, can be removed from the labeled context dataset and added to the single-class dataset. These steps can be repeated until no data points are classified by the model.