AI Model Personalization via Selective Training Data Generation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for updating artificial intelligence models require significant storage capacity, increased computer resources, and longer update times, and suffer from catastrophic forgetting when trained on new data without maintaining existing knowledge.
Innovation Solution
An electronic apparatus with a memory to store training data generation models and an AI model, which generates personal and general training data to train the AI model, allowing incremental updates and maintenance of user-specific characteristics without relying on vast external data storage or network connections.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If training data is continuously accumulated to improve AI model performance, then recognition accuracy is improved, but storage capacity requirements increase
Solution Approach 1:
The patent extracts only the essential training data characteristics needed for model improvement while discarding redundant data. The system identifies and retains only the most valuable training samples that contribute to recognition accuracy, thereby reducing storage requirements while maintaining model performance.
Solution Approach 2:
The patent changes the parameter of data representation by transforming raw training data into compressed feature vectors or embeddings. This parameter transformation allows the same information content to be stored with significantly reduced storage capacity while preserving the ability to train the AI model effectively.
2Measurement precision
If more training data is accumulated and the model is retrained, then model performance is improved, but the time required for model update increases
Solution Approach 1:
The patent applies partial action by training the AI model on only a selected subset of training data rather than the complete dataset. By identifying and using only the most relevant training samples, the system achieves model performance improvement without the computational time cost of processing all available data.
Solution Approach 2:
The patent performs preliminary action by pre-processing and selecting training data in advance before model training. The system prepares and organizes the most valuable training samples beforehand, which accelerates the actual model training process and reduces overall update time.
3Adaptability or versatility
If new training data is used to update the AI model, then model is updated with new information, but previously learned knowledge is forgotten (catastrophic forgetting)
Solution Approach 1:
The patent merges new training data with previously learned knowledge by combining new training samples with a curated subset of historical training data. This merging approach ensures that the AI model learns from new information while retaining previously acquired knowledge, preventing catastrophic forgetting.
Solution Approach 2:
The patent applies discarding and recovering by selectively discarding redundant new training data while recovering and preserving important previously learned knowledge. The system identifies which new data is truly necessary and maintains essential historical knowledge, achieving model updates without information loss.
Data Source
AI summary
Disclosed is an electronic apparatus. The electronic apparatus may include a memory configured to store one or more training data generation models and an artificial intelligence model, and a processor configured to generate personal training data that reflects a characteristic of a user using the one or more training data generation models, train the artificial intelligence model using the personal learning data as training data, and store the trained artificial intelligence model in the memory.