Combination of Neural Networks for Multi-label Document Classification

摘要: This paper deals with multi-label classification of Czech documents using several combinations neural networks. It is motivated by the assumption that different nets can keep some complementary information and it should be useful to combine them. The main contribution this consists in a comparison combination approaches improve results individual nets. We experimentally show all outperform nets, however they are comparable. However, best method supervised one which uses feed-forward net sigmoid activation function.