Siavash Haghiri

I am a PhD student in Theory of Machine Learning group (TML), at Tuebingen University, Computer science department under supervision of Prof. Dr. Ulrike von Luxburg. My main focus in my PhD program is on: ”Comparison-based methods in machine learning”. In traditional settings of machine learning, data is given either in the form of points in a Euclidean space or a matrix of distance/dissimarlity values between the pairs of points. However this is not always the case. Consider a recommendation system in an online movie database. To learn the preferences of users we need the dissimilarity (or similarity) of movies. This information is usually gathered from reviewers (users of the database) by asking them to provide answers to similarity questions among items. In this way, we ask the users to provide a similarity value in the range of 1 to 10 for pairs of items. These values highly depend on the users and can be biased. As an alternative approach one can ask if item A is more similar to item B or item C. Human beings are better in this kind of comparison tasks. We are concerned with developing algorithms, which take the comparisons and perform classic machine learning tasks, such as classification and clustering. Here you can find the related publications:

• S. Haghiri, D. Garreau, U. von Luxburg, “Comparison-Based Random Forests”, in ICML 2018. Paper Code
• S. Haghiri, D. Ghoshdastidar, U. von Luxburg, “Comparison-Based Nearest Neighbor Search”, in AISTATS 2017. Paper Code

• S. Haghiri, H. R. Rabiee, A. Soltani-Farani, S. A. Hosseini, M. Shadloo, "Locality Preserving Discriminative Dictionary Learning", IEEE Int. Conf. Image Processing, 2014.