Collaborative learning with taboos for machine learning methods in big data problems

The practical application of artificial intelligence methods has two big disadvantages. The first one is the amount of data needed to train models, and the other one is the lack of flexibility when changing data. In this paper, we propose an idea of collaborative learning for artificial intelligence methods with taboos which can be a solution for previously described problems. The main idea is to modify the first two rounds of collaborative learning solution for choosing the type of classifier and in the rest of them, the taboos lists are introduced. The classified data samples are added to the list and for some time are not used to focus on training data, where accuracy is lower. This novel architecture was described and analyzed using different machine learning approaches and big datasets for common classification problems.