MI (Multi-instance) learning is machine learning over MI data, where the instances are grouped together into labelled bags. An approach to handling MI data is propositionalisation, where each bag is converted into single feature vector, which can then be used with standard learning algorithms such as SVMs and Neural Networks.
The aim of this project is to explore adaptive propositionalisation, where the propositionalisation process adapts to the specific single instance learner being used. Application of this technique to the image classification problem will also be considered.