The optimization package quickly and efficiently solves a wide variety of linear, non-linear, and global optimization tasks.

Optimization Algorithms (Solvers) find the minimum, maximum, or zero of a function. EWA Systems' optimization package includes the full set of optimization algorithms: feasibility, constrained, unconstrained, and global optimization solvers for all of the relevant constraints types (Simple, Linear, Non-Linear) and objective function types (Linear, Quadratic, Non-Linear). Mixed-Integer solvers are also included. These solvers all implement common interfaces, enabling the user to easily switch out one algorithm for another. The package also includes a wizard algorithm that given an optimization problem, chooses the appropriate algorithm.

The optimization packages includes a number of algorithms, including: Revised Simplex, Newton, Quasi-Newton, and many problem-specific algorithms. Mixed Integer problems are handled with Branch and Bound. Global Optimization problems are handled with Genetic Algorithms. Many of these algorithms are multi-threaded and can be run on computer clusters.

These algorithms are optimized for speed and memory, performing well ahead of our competitor's algorithms. Problems that require over a GB of RAM with our competitor's algorithms are efficiently handled in just 80 MB. Problems that take 8 hours to solve with their algorithms, are solved in just 40 minutes.

Genetic Algorithms

EWA Systems implemented Genetic Algorithms as a complement to its Solver package in the field of Global Optimization problems. Global optimization problems are those problems where the problem space is chaotic, having many local maxima or minima, all of which need to be sampled in order to attempt to find the global maxima or minima. Genetic Algorithms are very effective in searching large problem spaces effectively.

EWA Systems' Genetic Algorithm Engine is a multi-threaded, multi-population that implements a wide variety of breeding and selection methods. The engine supports categorical, ordinal, and continuous variables and can be used to maximize, minimize or find zeros in the objective function. The following breeding and selection methods, and stopping conditons are supported. There are numerous example problems provided.

Breeding Methods Random Single Crossover Double Crossover Uniform Crossover Linear Recombinant Complete Recombinant Gradient Recombinant Mutation Migration

Selection Methods Best Tournament Linear-Rank Fitness-Rank Roulette

Stopping Conditions Max Generations Stagnation Solution Time

Example Genetic Algorithm Problems:

To Purchase or For More Information, contact our Sales Team.