Colloquia / Seminars / Industry Talks

This is the last seminar in a series of four seminars. The topic of this seminar will be a discussion on how relationships between atomic elements in an application domain can be predicted or estimated. For example, lets assume a new Web document is created and uploaded to a Web server. How can we predict which other Web document should link to this new page, and to which other page this new page should be linking to?
In fact, a growing number of areas in industry and research utilize relationships between data items in Data Mining applications. For example, recommender systems define relationships between users and objects which are exploited for the purpose of ranking objects subject to user preferences. Another example, Web search engines analyse the hyperlink structure of the Web to obtain an indication to the popularity of a Web document. Similarly, the scientific community relies heavily on the analysis of the inter-document reference (or citations) to obtain measures on the track record or author or the impact rating of conferences and journals.
A main problem with these domains is that dependencies between the objects are often user generated, and hence, are subject to spam, error, and abuse. Another main problem is known as the “rich-get-richer” problem which results in the application of many link analysis algorithms for the purpose or rating or ranking objects.
This seminar will introduce the audience into the world of interlinked objects, and the problems which we face in such domains. It will be shown that the domain and the associated problems require the addressing by us researchers since most people on this “blue marble” are affected. We will show that approaches can be developed which predict relationships between objects, and that these predictions can be used to verify the validity of existing links. This will be shown by using a large scale, real world application domain, and a machine learning approach to link prediction.

These are introductory seminars. Attendees of the seminars do not need to have prior knowledge in artificial neural systems.

Dr Markus Hagenbuchner will present the seminar. A case study will be presented by a guest speaker Dr Milly Kc from the University of Wollongong in Australia.

Markus Hagenbuchner holds a PhD (Computer Science, University of Wollongong, Australia). He is currently a senior lecturer in the School of Computer Science and Software Engineering at the University of Wollongong, Australia. He joint the machine learning research area in 1992, started to focus his research activities on Neural Networks for the graph structured domain in 1998, and pioneered the development of Self-Organizing Maps for structured data. His contribution to the development of a Self-Organizing Map for graphs led to winning the international competition on document mining on several occasions.

He is a team leader of the machine learning group at the University of Wollongong. He has been the co-chair for the AI-08 conference, and is a program committee member for ANNPR 2010. He has been a reviewer of international standing for the Australian Research Council since 2004, and has been an invited guest speaker at various international venues. His current research interest is on the development of supervised and unsupervised machine learning methods for the processing of complex data structures in data mining applications.