Execution-Profile Matching for Machine Learning Model Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing AI systems lack a unified format for providing information such as input signal type, calculation amount, detection result frequency, and output coincidence with user needs, requiring manual operation to select machine learning programs suitable for a specific execution environment, hindering market circulation.
Innovation Solution
An information processing apparatus and method that includes a storage unit for machine learning models and profile information, enabling automatic extraction of models executable in a user-defined environment based on profile information, and automatic setting of input and output formats.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual operation is performed to select machine learning programs by checking detailed descriptions, then the user can identify suitable programs for their execution environment, but the process becomes time-consuming and complex
Solution Approach 1:
The patent applies preliminary action by pre-creating profile information that describes execution conditions (input signal types, calculation amounts, detection frequencies, output formats) for each machine learning program. This profile information is prepared in advance and stored in a database, allowing users to quickly query and select suitable programs without manually checking detailed descriptions, thus resolving the contradiction between selection accuracy and time consumption
Solution Approach 2:
The patent introduces profile information as an intermediary between machine learning programs and users. This profile information acts as a mediator that translates complex program characteristics into standardized, queryable formats, enabling efficient matching between user requirements and program capabilities without direct manual inspection of program details
2Reliability
If manual checking of detailed descriptions is performed for each program, then the user can determine operability and detection result output, but the operation complexity increases
Solution Approach 1:
The patent applies segmentation by dividing program characteristics into distinct, standardized profile information elements (input signal type, calculation amount, detection frequency, output format). This segmentation allows users to check specific aspects independently through structured queries rather than reviewing comprehensive detailed descriptions, reducing operational complexity while maintaining reliability
Solution Approach 2:
The patent transforms program characteristics into standardized parameters within profile information (e.g., calculation amount as a quantifiable parameter, detection frequency as a rate parameter). This parameterization enables systematic comparison and filtering of programs based on user requirements, simplifying the selection process while ensuring reliable operability confirmation
3Ease of operation
If profile information is added to AI logic, then the user can easily select needed machine learning models, but information storage requirements increase
Solution Approach 1:
The patent applies universality by creating a standardized profile information structure that serves multiple functions: it describes execution conditions, enables query-based selection, supports automated matching, and facilitates market circulation of machine learning programs. This multi-functional profile information reduces the need for separate documentation systems, making the increased storage requirement more efficient and justifiable
Data Source
AI summary
A server device stores therein a plurality of machine learning models and profile information that indicates an execution condition of each of the machine learning models, and extracts, when receiving a request from an operator, machine learning models that are executable in an execution environment that is requested by the user from among the plurality of machine learning models based on the profile information.