OASIS OPTIMIZATION ALGORITHMS
OASIS proprietary optimization algorithms currently include:
- SOGO (Single Objective Global Optimization): solving global optimization problems with one objective and many inexpensive constraints
- MOGO (Multi-Objective Global Optimization): solving global optimization problems with more than one objective with inexpensive constraints
- SOGO-C (SOGO for Constrained Problems): solving global optimization problems with expensive constraints and/or tightly-constrained search spaces
Built upon more than 20 years of university research, these algorithms intelligently integrate methods from machine learning, statistics, and mathematics to efficiently and effectively search for the best design solution.
The algorithms apply intelligent Design of Experiments (DOE) or sampling to generate exploratory design points, drive the computer analysis/simulation or physical experiments. The results are then processed using a combination of metamodeling, machine learning, statistical analysis, and mathematical programming approaches, in order to intelligently determine where to generate the next batch of samples. This process iterates until convergence.
Features of OASIS Algorithms
All three OASIS algorithms share common features:
- Solution for linear/nonlinear, discrete/continuous, and unimodal/multimodal problems.
- Superior performance from low-scale (number of variables less than 10) to large scale problems
- Direct integration with external analysis or simulation, no equations necessary
- No algorithm picking
- No algorithm parameter tuning
- Effective optimization with fewer simulation calls