Why globalMOO? A Technical Overview
globalMOO is a breakthrough technology for multi-objective optimization and the efficient solution of inverse problems. Existing numerical approaches handle single-output inverse models, but no efficient methods exist for multi-objective inverse solutions.
The Problem with Current Approaches
Heuristic Methods
Tailored to specific problem descriptions. Not generalizable to arbitrary multi-objective problems.
Scalarization
Lumps multiple outputs into one proxy via loss function, producing biased solutions that don't represent multidimensional trade-offs.
Evolutionary / Pareto
Solves multiple objectives truly but requires astronomical compute resources and excessive model evaluations.
globalMOO solves for multiple objectives efficiently, even if the objectives are in different units and scales, without requiring human intervention.
What is globalMOO?
An expert system that captures knowledge related to a software algorithm and uses that information for the inverse solution of that algorithm. It is model agnostic — works with physics-based models, ML models, and others.
Inverse solution: output values are specified (objectives), and corresponding inputs are back-calculated. Knowledge capture leads to understanding cause-and-effect relationships inherent in the forward model.
Where is globalMOO Applicable?
Compatible with spreadsheets, physics-based simulators, analytical equations, and AI regression/correlation models. Extends to hundreds of input variables and thousands of objectives.
Model Requirements
Input variables and outputs can be continuous, integer, logical, or categorical. Outputs must be precisely (non-stochastically) computed from inputs. The model must be causally closed and input-complete.
Flexibility
Forward model need not be continuous or differentiable. Multiple non-unique solutions can be found. Requires valid solutions in at least 90% of the search space.
Two Applications
Optimization
Match, minimize, and/or maximize multiple objectives simultaneously.
Impact Factor Calculation
Quantify how much each input variable influences outcomes — calculate the bias of the algorithm with respect to each variable.
