Ontology-Guided ML Training With LLM Model and Data Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current processes for training machine learning models are complex, unintegrated, and time-consuming, making them less accessible to users.
Innovation Solution
The system employs an ontology and Large Language Models (LLMs) to automatically select and prepare training data, choose machine learning model types, set up parameters and configurations, and deploy customized models to perform specific tasks, with an AI agent supervising the process to avoid errors and select the best model.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If manual processes are used for training machine learning models (identifying training data, fine-tuning settings, configuring parameters), then users can control and understand each step, but the process becomes complex, time-consuming, and less accessible
Solution Approach 1:
The system enables self-service automation where the ML model training process automatically identifies relevant training data, selects appropriate model types, configures parameters, and fine-tunes settings without requiring user expertise. The automation service performs these tasks independently based on user requests, transforming a complex manual process into an accessible automated service.
Solution Approach 2:
An automation service acts as an intermediary between the user and the complex ML training infrastructure. This intermediary layer handles the complexity of data identification, model selection, parameter configuration, and training execution, while presenting a simplified interface to users. The intermediary translates user requests into detailed training workflows without exposing users to underlying complexity.
2Productivity
If automated processes are implemented for selecting training data and configuring models, then the process becomes faster and more accessible, but the system complexity increases
Solution Approach 1:
The automation service provides multi-functional capabilities by combining multiple tasks into a single unified service: training data identification, ML model type selection, parameter configuration, and training execution. This universal service handles diverse ML training requirements through a single interface, increasing productivity while managing system complexity through consolidation rather than proliferation of separate components.
Solution Approach 2:
The system performs preliminary actions by pre-identifying relevant training data and pre-selecting appropriate model types before the actual training process begins. This advance preparation automates time-consuming tasks that would otherwise require user intervention during the training workflow, thereby increasing deployment speed while the complexity is managed through systematic pre-processing.
3Measurement precision
If users manually identify relevant training data and fine-tune parameters, then model accuracy can be optimized, but the process becomes time-consuming and requires sophisticated expertise
Solution Approach 1:
The system replaces the mechanical process of manual data identification and parameter fine-tuning with automated computational processes. The automation service uses algorithms and ML techniques to identify relevant training data and optimize parameters, substituting human expertise with systematic automated analysis. This maintains model accuracy while dramatically reducing the time required, as automated processes can analyze and optimize parameters much faster than manual processes.
Solution Approach 2:
The automation service implements feedback mechanisms during the training process to continuously monitor model performance and automatically adjust parameters for optimization. This automated feedback loop replaces manual fine-tuning by systematically evaluating model outputs and making real-time parameter adjustments, maintaining high accuracy while reducing the time users would need to spend on iterative refinement.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Computer-implemented systems and methods including language models for explaining and resolving code errors. A computer-implemented method may include: receiving one or more user inputs identifying a data set and providing a first user request to perform a first task based on at least a portion of the data set, wherein the data set is defined by an ontology; using a large language model ("LLM") to identify a first machine learning ("ML") model type from a plurality of ML model types; using the LLM to identify a first portion of the data set to be used to perform the first task; using the LLM to generate a first ML model training configuration; and executing the first ML model training configuration to train a first custom ML model, of the first ML model type, to perform the first task.