Learning Machines Information Distribution System with Example Applications

When problem solving reduces to examination of a single or a few learning methods no sophisticated mechanisms of information exchange are necessary, but when we use meta-learning for extensive search through a huge space of hybrid models, the information exchange between subsequent models is crucial. The information exchange between models must be universal, very flexible and as simple to define as possible. The design of an efficient system must include abstract methodology for transmission of amorphous information between different kinds of methods and optimizing different types of functions. It is highly important for meta-learning where different types of information must be collected and used by meta-processes at high level of abstraction. This article presents a universal information exchange system eligible for huge data mining tasks which was implemented in our meta-learning environment Intemi.