Optimal Design

Optimal Design specializes in development of high added-value optimization software. Most optimization problems being addressed by our products are very difficult to solve, due to the huge size and complexity of its data, requiring innovative advanced technology to deliver reliable and high quality solution. Several advanced optimization techniques are exploited in our products, each time precisely targeted to the particular optimization problem being tackled. The technology of genetic algorithms is a core foundation of every product we design and develop.

More than a decade of top-level research into genetic algorithms (see publications) allowed us to become a leader in that domain, building a proprietary cross-platform library exploited in many of our software.

Of special interest is the Grouping Genetic Algorithm, a flavor specifically designed for clustering problems that was invented by Optimal Design.

The following is a short introduction to the basic technique of genetic algorithms.

Genetic Algorithms

    Disclamer intended for geneticists
The following explanations illustrate the basic ideas behind genetic algorithms. Although the technique is inspired by natural processes, the natural phenomena have been drastically simplified in the technique. We are well aware that true "wetware" genetics is incomparably more complex.

The genetic algorithm is an optimization technique introduced by John Holland in 1975. The principle of this technique is to simulate the evolution in nature.

Obtaining a good solution in complex problems can be very difficult because your computer is not fast enough to try all the possibilities. The genetic algorithm generates a pool of solutions, some will be better than the others and the cross of these solutions could yield a better one still.

If we take the case of the animal world, we can consider one chromosome of a particular beast as one possible solution for the species. We will see how we can create a more powerful animal from two who had some good characteristics.

A simple chromosome

From the group of beasts we had, we will select one with a good sight and one with strong feet. As seen in the following figure (Classic crossover) these two parents could generate two offspring, one would be better than either of the two creators, and one will probably have no particular charateristics.

Classic crossover


After cross


We have now generated a better animal. The main idea of the genetic algorithm is to cross this new animal again with others, always attempting to create better creatures.

The above is the principle of the classic genetic algorithm. Different real-world optimization problems require different enhancements to the basic technique. Optimal Design's expertise with the technique allows us to design the proper enhancement for each target problem. For instance, our proprietary grouping genetic algorithm is well suited for grouping problems.

Details of our methods can be found in the book of our co-founder Dr. E.Falkenauer.