Data Augmentation Selection for Cognitive Systems
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current data augmentation methods for cognitive systems are ad hoc and lack a systematic approach to determine the optimum augmentation methods for a specific system, leading to inconsistent improvements in accuracy across different cognitive systems.
Innovation Solution
A method that applies a set of data augmentation approaches with varying parameters, computes goodness scores for each augmented model through a set of features, and ranks them to select the best model, allowing for automated selection of the optimum data augmentation method.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If ad hoc data augmentation methods are used, then the training process can be simplified, but the accuracy improvement becomes inconsistent and unreliable
Solution Approach 1:
The patent systematically varies multiple parameters of data augmentation methods (augmentation type, intensity, combination strategies) to create a comprehensive search space. This allows automated identification of optimal parameters that reliably improve accuracy while maintaining training simplicity through parameter optimization rather than complex manual tuning.
Solution Approach 2:
The patent implements a feedback loop where validation scores from trained models are used to evaluate and rank different data augmentation methods. This feedback mechanism enables automatic selection of the best augmentation methods based on actual performance improvement, ensuring reliable and consistent accuracy enhancement across different cognitive systems.
2Adaptability or versatility
If multiple data augmentation methods are applied, then the training data diversity increases, but the system complexity increases
Solution Approach 1:
The patent segments the data augmentation process into distinct, independently evaluable methods (e.g., synonym replacement, back-translation, paraphrasing). Each method can be applied separately and evaluated independently, allowing the system to manage complexity through modularization while maintaining high data diversity through combination of multiple segmented approaches.
Solution Approach 2:
The patent creates a universal evaluation framework that can assess multiple different data augmentation methods using the same validation criteria and scoring mechanism. This universal approach allows diverse augmentation techniques to be systematically compared and selected based on their performance improvement, reducing the need for system-specific complex tuning for each method.
3Productivity
If automated selection of augmentation methods is implemented, then the efficiency improves, but the validation and ranking process becomes more complex
Solution Approach 1:
The patent implements self-service through automated evaluation where the system automatically applies data augmentation methods, trains models, validates performance, and ranks methods without requiring manual intervention. The validation framework automatically computes improvement scores based on predefined criteria, enabling efficient automated selection while managing complexity through standardized, reusable validation routines that can be applied consistently across different methods.
Data Source
AI summary
A cognitive system (artificial intelligence) is optimized by assessing different data augmentation methods used to augment training data, and then training the system using a training set augmented by the best identified method. The augmentation methods are assessed by applying them to the same set of training data to generate different augmented training data sets. Respective instances of the cognitive system are trained with the augmented sets, and each instance is subjected to validation testing to assess its goodness. The validation testing can include multiple validation tests leading to component scores, and a combined validation score is computed as a weighted average of the component scores using respective weights for each validation test. The augmentation method corresponding to the instance having the highest combined validation score is selected as the optimum augmentation method for the particular cognitive system at hand.