ML-Based Optimization Algorithm Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional optimization algorithms for complex problems like supply chain management require extensive computational resources and time, as each algorithm takes varying amounts of time to execute and produces solutions of different quality, making it inefficient to apply all algorithms to find the best solution.
Innovation Solution
Implementing machine learning systems that select the most appropriate optimization algorithm in real-time based on problem features, such as variables and constraints, to predict run-time and quality metrics, thereby reducing the need to execute multiple algorithms and optimize resource usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If all optimization algorithms are executed to find the best solution, then solution quality is improved, but CPU time and computational resources are excessively consumed
Solution Approach 1:
The system performs preliminary classification of the optimization problem using machine learning models before executing any optimization algorithms. This preliminary action identifies the most suitable algorithm family based on problem features, preventing unnecessary execution of incompatible algorithms and reducing overall CPU time while maintaining solution quality.
Solution Approach 2:
The system uses self-learning machine learning models that automatically adapt to different optimization problem types. The models are trained on historical data and continuously improve their ability to predict the best algorithm selection, enabling the system to serve itself without manual intervention and optimize the balance between solution quality and computational time.
2Measurement precision
If multiple optimization algorithms with different options are tried to find the best solution, then solution quality is improved, but device complexity increases
Solution Approach 1:
The system segments the optimization algorithm portfolio into distinct families or categories based on problem characteristics. Machine learning models classify problems into specific segments and select only the relevant algorithm families for execution, reducing the effective complexity from the full portfolio while maintaining the ability to find high-quality solutions for each problem type.
Solution Approach 2:
The machine learning model acts as an intermediary between the problem definition and the optimization algorithm portfolio. It translates problem features into algorithm selections, simplifying the interaction complexity and providing a systematic approach to navigating the algorithm portfolio without manual intervention.
3Measurement precision
If all optimization algorithms are executed on complex supply chain problems, then solution accuracy is improved, but data storage and computational infrastructure requirements increase
Solution Approach 1:
The system performs preliminary classification of supply chain optimization problems using machine learning models before executing any optimization algorithms. This preliminary action identifies the most suitable algorithm family based on problem features, preventing unnecessary execution of incompatible algorithms and reducing overall CPU time while maintaining solution quality.
Solution Approach 2:
The system uses self-learning machine learning models that automatically adapt to different optimization problem types. The models are trained on historical data and continuously improve their ability to predict the best algorithm selection, enabling the system to serve itself without manual intervention and optimize the balance between solution quality and computational time.
Data Source
AI summary
Systems and methods that extract features from a set of optimization problems, and compile performance characteristics of optimization algorithms that are applied to each optimization problem. Machine learning models are trained on a first portion of a dataset that comprises the features and performance characteristics. A model is selected based on performance on a second portion of the dataset. The selected model is applied to features of a new optimization problem to provide performance characteristics of each optimization algorithm, which can then be ranked based on the respective performance characteristics. Either the first-ranked optimization algorithm can be applied to the new optimization problem, or successively-ranked optimization algorithms can be executive iteratively.