Whatever activity we pick in the course of building a model from empirical data or maintaining an efficient knowledge base it will finally lead to some kind of an optimization problem. Thus it is necessary to elect useful subsets of factors and base functions, optimize structure of mathematical model, find numeric coefficients, etc. For all these tasks there are no existing standard methods, especially in the case of high number of dimension of the input vector, which is consistent with most of real world learning from data tasks.

Thus existence of a reliable and robust optimization technique capable to effectively work in different types of search space is critical to success of our research. In our developments this role is given to Evolutionary Search - an object oriented version of Genetic Algorithm. One of the most important advantages of Evolutionary Search is its efficiency in such tricky non-topological search spaces like spaces of tree or list structures.

Demo on Evolutionary Search can give you general idea of the method and evaluate its capacity for work: 

Despite the fact that Evolutionary Search can be useful in a lot of practical tasks like resources distribution, building schedules, operations research, etc., we believe that it has great undiscovered potential, that we are going to research and make available. As regards improvement of the Evolutionary Search we particularly rely on self-learning, which might be able to generalize different search situations and suggest optimal search parameters for a particular new situation or even preferred direction in the search space.