Generativní a diskriminativní modely pro množinová data

Abstract

This research project deals with hybrid discriminative and generative modeling and its possible utilisation in multiple-instance learning. At first, technicalities of this approach are introduced, consequently a simple experiment is performed and lastly multi-instance laerning itself is taken care of. The main aim of this work is to use the HMill framework with the Mili.jl library and involve the contrastive learning into it, where benefits of this approach are shown