Information Processing for Visual Machine Learning Model Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems struggle to assist users in selecting a suitable machine learning model from a large number of models due to the lack of effective methods for evaluating and displaying model performance.
Innovation Solution
An information processing system that manages model information associated with favorable and unfavorable processing targets, generating display control information to showcase these targets alongside model information, facilitating user selection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If only model output and determination results are displayed in a table, then the display format is simple, but it becomes difficult for users to select suitable models from a large number of machine learning models
Solution Approach 1:
The patent transitions from a two-dimensional table display to a three-dimensional visual display by adding image dimensions. Model outputs are displayed as images with overlay markings showing favorable and unfavorable regions, creating a spatial visualization that preserves detailed performance information while improving accessibility and ease of comparison across multiple models.
Solution Approach 2:
The patent creates visual copies of model outputs in the form of displayed images with overlay markings. Instead of merely tabulating results, the system generates visual representations that copy the essential performance characteristics of each model, allowing users to intuitively compare model behaviors across different inputs and scenarios.
2Measurement precision
If detailed model evaluation information is provided for all models, then model selection accuracy improves, but the processing load on client devices increases
Solution Approach 1:
The patent extracts only the essential evaluation information needed for model selection - specifically, representative images with overlay markings showing favorable and unfavorable regions. By taking out only these critical visual elements rather than transmitting complete model datasets or detailed numerical evaluations, the system maintains high evaluation precision while significantly reducing the data volume and processing energy required on client devices.
Solution Approach 2:
Instead of having client devices process and analyze complete model datasets to evaluate performance, the patent inverts the approach by having the server pre-process models and generate condensed visual evaluation results. The client then only needs to display and compare these pre-processed images, reversing the computational burden from client to server while maintaining evaluation precision.
3Ease of operation
If visual display of model outputs is implemented, then model comparison becomes more intuitive, but the data transmission volume increases
Solution Approach 1:
The patent creates condensed visual copies of model outputs that capture essential performance characteristics in a single image with overlay markings. Rather than transmitting multiple raw data points or complete datasets, the system generates representative visual copies that convey model behavior intuitively while occupying minimal transmission bandwidth.
Solution Approach 2:
The patent transforms model evaluation data from numerical parameters to visual parameters. By changing the representation format from tables of numbers to images with spatial markings, the system enables more intuitive comparison while actually reducing the effective information volume, as human visual processing can comprehend complex model behaviors in a single image rather than requiring extensive numerical data.
Data Source
AI summary
An information processing device comprises a management unit configured to manage database information in which model information regarding a learned model is associated with at least one of a favorable item and an unfavorable item among processing targets processed by the learned model, the learned model being good at processing the favorable item and not good at processing the unfavorable item; and a generating unit configured to generate display control information for displaying the model information and information regarding at least one of the favorable item and the unfavorable item associated with the model information based on the database information.