Hierarchical AI Experiment History Interface for Model Tracking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing artificial intelligence (AI) model development processes face challenges in efficiently managing experiment results and histories, leading to resource wastage and difficulty in tracking experiment settings.
Innovation Solution
A user interface with a hierarchical structure is provided, allowing for the display and management of AI model experiment history and enabling user interactions to perform and update experiments, including training, compression, and benchmarking, with efficient display and storage of experiment results and histories.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If numerous experiments are conducted with various hyperparameters, datasets, and algorithms, then the quality and comprehensiveness of AI model development improves, but the difficulty of tracking results and settings increases
Solution Approach 1:
The system segments experiment management into distinct components: experiment definitions (with hyperparameters, datasets, algorithms), experiment results (performance metrics), and experiment histories (timeline of executions). This segmentation allows each component to be managed independently while maintaining comprehensive tracking through the hierarchical structure.
Solution Approach 2:
The patent introduces a temporal dimension through experiment histories that display experiments chronologically. The hierarchical structure adds organizational dimensions by grouping experiments under models and submodels, transforming the flat complexity into multi-dimensional organized data that is easier to navigate and track.
2Productivity
If more experiments are conducted, then the optimization capability of the AI model improves, but the time and resources required for tracking and managing experiments increase
Solution Approach 1:
The system performs preliminary organization by defining experiment templates with all necessary parameters (hyperparameters, datasets, algorithms) before execution. Experiment histories are automatically generated and updated as experiments are conducted, eliminating the need for manual tracking after experiments are completed.
Solution Approach 2:
The hierarchical structure creates organized copies of experiment data at different levels (models, submodels, experiments). Each experiment result is automatically recorded and linked to its parent model, allowing comprehensive tracking without manual intervention. The system copies and stores relevant information in an organized manner that enables efficient retrieval and analysis.
3Ease of operation
If a hierarchical structure with multiple experiment categories is implemented, then the organization and visualization of experiment data improves, but the complexity of the user interface increases
Solution Approach 1:
The user interface is segmented into distinct functional areas: a first area displaying the hierarchical structure with experiment categories and histories, and a second area displaying detailed experiment results. This segmentation allows complex information to be presented in organized, manageable sections rather than overwhelming the user with all data simultaneously.
Solution Approach 2:
The hierarchical structure organizes data across multiple dimensions (models, submodels, experiments, categories) that can be navigated systematically. The interface presents this multi-dimensional data in a structured format that users can explore at different levels of detail, transforming complexity into navigable organizational layers.
Data Source
AI summary
In accordance with an embodiment of the present disclosure, a method performed by a computing device is disclosed. The method includes presenting a user interface for a development project of an artificial intelligence model The method includes in response to receiving an object selection input to select a first lower object connected dependently to a first upper object in the hierarchical structure of the first area. The method includes displaying a first experiment of a first model corresponding to the first lower object in the second area. The method includes in response to receiving an experiment input to perform a second experiment of the first model in the second area, displaying information related to the second experiment in the second area and updating the hierarchical structure in the first area based on the second experiment.