Next Level of Decision Intelligence
Agent Driven Multi-Objective Optimization
globalMOO revolutionizes optimization across industries, delivering unmatched speed and efficiency.
WHAT is globalMOO? - Agent-Based Expert System for Inverse Solutions
globalMOO is a software API designed to optimize multiple objectives using many agents that are self-learning. We deliver a DLL with a powerful and easy-to-integrate API in multiple programming languages. We also provide an option for web API integration. Easy plug-in, only five required endpoints for integration.
Distinguishing Features
- globalMOO solves for multiple objectives efficiently using a multithreaded parallel implementation, even if the objectives are in different units and scales, without requiring human intervention. No subjective weights.
- globalMOO can solve the inverse of systems of linear and non-linear complex problems. No need to simplify your models.
- globalMOO can minimize one objective, maximize another, match the defined targets for others and constrain (greater than and/or less than) other targets all at the same time. KKT or Pareto optimization in one pass.
- globalMOO can be applied to any predictive algorithm, be it a spreadsheet, physics-based simulator, system of analytical equations, or an AI-based model. Model and platform agnostic. No need to reveal your models.
- globalMOO can handle, float, integer and logical variables, in combination with categorical variables. Flexible, applicable to a wide range of problems without modifications.
- globalMOO can detect insensitive variables and allow the user to re-scope the problem such that these variables or variable-ranges are excluded. Explainability.
- globalMOO can find multiple non-unique inverse solutions for a given set of outcomes, allowing users to better understand and characterize the non-uniqueness inherent in their application. Model Auditing for Adequacy and Trustworthiness.
No subjective weights.
- globalMOO solves for multiple objectives efficiently using a multithreaded parallel implementation, even if the objectives are in different units and scales, without requiring human intervention.
KKT or Pareto optimization in one pass.
- globalMOO can minimize one objective, maximize another, match the defined targets for others and constrain (greater than and/or less than) other targets all at the same time.
Model agnostic. No need to expose, simplify or reformulate your models.
- globalMOO can be applied to any predictive algorithm, be it a spreadsheet, physics-based simulator, system of analytical equations, or an AI-based model.
- globalMOO can solve the inverse of black-box models, including linear and non-linear complex problems.
Flexible, applicable to a wide range of problems without modifications.
- globalMOO can handle, float, integer and logical variables, in combination with categorical variables.
HOW does it work? - Just like subject matter experts, globalMOO agents learn from mistakes and gain experience
In the learning stage, globalMOO prescribes a limited number of predictive model runs (numerical experiments) for its Agents to train on the inverse solution. Later, in the application stage, globalMOO provides guidance on how to update the input variables to reach the desired objectives.
WHY use globalMOO? - True Multi-Objective Optimization
- globalMOO performs targeted optimization using inverse solution from the desired system objectives which speeds up the solution time and more efficiently uses computational resources. Energy efficient.
- Approaches that rely on heuristic methods must be tailored for the specific problem at hand. globalMOO does not rely on subjective judgements, therefore it is applicable to any predictive model with precise outcomes. No user intervention.
- Many predictive software models yield outputs that are in different units. Current approaches for multi-objective optimization of such models require scalarization through utility functions with heuristic weight assignments. Hence, the results are user-dependent and can rarely be described as optimal. globalMOO does not use heuristic weights (user-agnostic), hence it provides user-independent results. No scalarization.
- For Pareto-optimal search and selection, existing approaches require astronomical compute resources and excessive numbers of model evaluations. globalMOO finds the Pareto-optimal conditions with a greedy learning algorithm. Fast and reliable.
- For Karush-Kuhn-Tucker (KKT-optimal) conditions, there are no known numerical approaches for larger problem sizes. globalMOO finds the KKT-optimal solution with a greedy learning algorithm for fast and reliable application. KKT-optimality.
WHERE can it be used? - Practical Applications
Calibration: All predictive models require calibration to the historically observed data. globalMOO’s inverse solution capabilities for multiple outputs (including temporal) enables the user to define and solve for many variables to match the data.
Optimization: globalMOO uses inverse solutions to minimize and/or maximize multiple objectives iteratively, including Pareto or KKT optimization.
Calculation of impact factor: Multiple globalMOO inverse solutions can be used to quantify the impact factor of input variables upon a particular set of outcomes. This calculates the bias of the algorithm for each variable.
WHO? - Anybody with a Predictive Model to Solve
Possible Applications
Civil and Mechanical
- Structural optimization
- Robotics
Petrochemical and Industrial
- Process design
- Manufacturing
- Drilling and completion
Electrical
- Circuit design
- Wireless networks
- Microgrids
Finance and Macroeconomics
- Monetary policy
- Portfolio management
Artificial Intelligence
- Hyperparameter optimization
- Explainability