Fast heterogeneous boosting

The main goal of this paper is introduction of fast heterogeneous boosting algorithm. `Heterogeneous´ means that boosting is based not on single-type learning machine, but may use machines of several types coherently. The main idea behind the construction of heterogeneous boostings was to use it with learning machines of low complexity (O(nd)). Thanks to that, the heterogeneous boosting is still a fast algorithm of linear learning (and usage) complexity. The paper presents a comparison of homogeneous boostings of a few types of fast learning machines with introduced heterogeneous boosting, which base on a small group of fast learning machines. The presented comparison proves that heterogeneous boosting is efficient and accurate.