ML Root Model Format Conversion for Continuous Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning models without pipeline capabilities face challenges in seamless continuous learning due to incompatibilities with pipeline-enabled models, limiting their deployment and updating in systems with limited computing power.
Innovation Solution
A computer-implemented method and system for converting a machine learning root model format into a target format without pipelines, using a learning system to generate a new version of the target model, determining quality threshold values via X control chart methods, and evaluating the quality metric values to ensure the target model meets the quality standards of the source model.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If machine learning models use pipeline architecture to enable continuous learning, then the continuous learning capability is improved, but the device complexity and computing power requirements increase
Solution Approach 1:
The patent extracts the pipeline architecture from the machine learning model, separating the data processing workflow from the model training process. This allows models without pipeline capabilities to benefit from continuous learning by processing data through external pipeline structures while maintaining their simpler architecture.
Solution Approach 2:
The patent introduces an intermediary data processing layer that acts as a mediator between the simple ML model and the continuous learning requirements. This intermediary handles the pipeline operations, allowing the core model to remain simple while still enabling continuous learning through the mediating structure.
2Device complexity
If machine learning models without pipeline capabilities are used to reduce device complexity, then the ease of deployment is improved, but the continuous learning capability deteriorates
Solution Approach 1:
The patent segments the continuous learning functionality into separate components: the simple ML model remains independent and simple, while the pipeline architecture is segmented into distinct data processing stages. This segmentation allows the simple model to be deployed easily while the segmented pipeline components provide the continuous learning capability externally.
3Manufacturing precision
If quality threshold validation is performed using X control chart methods to ensure model quality, then the manufacturing precision is improved, but the productivity and deployment speed may be reduced
Solution Approach 1:
The patent applies preliminary action by performing quality validation using X control chart methods before model deployment. This preliminary quality check ensures that only models meeting the quality thresholds are deployed, preventing poor-quality models from entering production and reducing the need for corrective actions later.
Solution Approach 2:
The patent implements feedback mechanisms through the X control chart methodology, where quality metrics are continuously monitored and fed back into the model development process. This feedback loop ensures that quality standards are maintained while allowing for iterative improvements without compromising deployment speed.
Data Source
AI summary
An ML root model format having a root model definition, is converted into an ML target model format not having the root model definition. A learning system is assigned to the root model definition that is convertible to the machine learning target model format. A new version of the root model definition is ported from the ML root model to the format of the ML target model to generate a new version of the machine learning target model after a learning iteration of the learning system. Quality limits are determined using an X control chart method based on a cross-validation of fold results generated during a validation of the new version of the machine learning root model. A quality metric value of the new version of the ML target model is evaluated against the quality threshold values of the new version of the ML root model.