Gossip Learning on Social Networks

The project is focused on building fully distributed algorithms to solve classification problems in machine learning. It will be based on the work done by Mark Jelasity's group (see papers below). The aim is to to extend this work to Social Network environments and analyse the behaviour of the algorithms when only local interactions are permitted on the social graph.

• Róbert Ormándi, István Hegedűs, and Márk Jelasity, "Asynchronous peer-to-peer data mining with stochastic gradient descent", in Emmanuel Jeannot, Raymond Namyst, and Jean Roman, editors, Euro-Par 2011, volume 6852 of Lecture Notes in Computer Science, pages 528–540. Springer-Verlag, 2011.
• Róbert Ormándi, István Hegedűs, and Márk Jelasity. Gossip learning with linear models on fully distributed data. Concurrency and Computation: Practice and Experience, 2012. to appear.
• István Hegedűs, Róbert Busa-Fekete, Róbert Ormándi, Márk Jelasity, and Balázs Kégl. Peer-to-peer multi-class boosting. In Euro-Par 2012, Lecture Notes in Computer Science. Springer-Verlag, 2012. to appear.