Hierarchical Multi-Armed Bandit for AI Model Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Selecting the appropriate artificial intelligence (AI) model for lexicon expansion is challenging due to the large number of available models, with some being off-topic or requiring priming with samples or feedback, and existing techniques struggle to incorporate subject-matter expert feedback and handle multiple models effectively.
Innovation Solution
A hierarchical multi-armed bandit method (MArBLE) is used to select AI models based on user feedback, employing frequency-based and history-based boosting strategies to alleviate the cold-start problem and dynamically improve model selection across multiple models without prior knowledge of architectures or training data distributions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If multiple AI models are used for lexicon expansion, then the coverage and diversity of candidate entities improve, but the complexity of model selection and management increases
Solution Approach 1:
The system implements self-service through automatic model selection mechanisms that evaluate and choose appropriate AI models based on task characteristics and performance metrics, eliminating the need for manual model selection and management by users
Solution Approach 2:
The framework provides universality by designing a unified model selection architecture that can handle multiple AI models with different architectures and training data distributions through a single interface, making the system adaptable to various lexicon expansion tasks
2Measurement precision
If subject-matter expert feedback is incorporated, then the accuracy of model selection improves, but the time and resources required for validation increase
Solution Approach 1:
The system applies partial action by selectively incorporating subject-matter expert feedback only for critical validation stages and using automated evaluation for routine assessments, thereby maintaining accuracy while reducing overall time and resource requirements
Solution Approach 2:
The framework implements feedback mechanisms where subject-matter expert evaluations are systematically collected and used to refine model selection criteria, creating a continuous improvement loop that enhances accuracy over time
3Ease of operation
If models are selected without prior knowledge of architectures or training data, then the ease of operation improves, but the reliability of model selection may deteriorate
Solution Approach 1:
The system introduces an intermediary component that automatically analyzes model architectures and training data characteristics, serving as a mediator between the user's lack of prior knowledge and the reliable selection of appropriate models based on objective criteria
Data Source
AI summary
Disclosed herein are methods, systems, and computer program products for selecting an artificial intelligence (AI) model. Aspects include receiving, by a multi-armed bandit agent, candidate entities from multiple set expansion models and selecting a subset of the candidate entities for a first set expansion task, wherein a first candidate entity of the selected subset of candidate entities is selected from a first model by a subject matter expert. Aspects also include selecting, by the multi-armed bandit agent, based on the first candidate entity selected by the subject matter expert, the first model from the set expansion models for generating further candidate entities.