Marius Kloft is a computer scientist with core expertise in statistical machine learning. His interests lie in virtually all aspects of ML, including theoretical analysis, development of novel methods, and their application for statistical data analysis. He is an expert in multi-task and multiple kernel learning (organized workshops at NIPS ’10, ’13’,’ 14; Dagstuhl ’15) and is currently co-organizing the only journal special issue on the topic of multi-task learning (in the premier Journal of Machine Learning Research). His collaborations with colleagues in interdisciplinary projects in computer security, computer vision, and the bio-medical domain have led to award-winning new methods (including the most accurate transcription start site finder and an ImageCLEF2011 challenge winning).
Staff in PREDICT:
Dr. Patrick Jähnichen
Drug prediction, model development and tuning, text mining support
Most Relevant Publications:
Cortes, M. Kloft, M. Mohri (2013). Learning Kernels Using Local Rademacher Complexity. Advances in Neural Information Processing Systems 26: 2760-2768 (awarded with Google Most Influential Paper Award).
M. Kloft and P. Laskov (2012). Security Analysis of Online Centroid Anomaly Detection. Journal of Machine Learning Research, 13(Dec):3647-3690.
M. Kloft and G. Blanchard (2012). On the Convergence Rate of Lp-Norm Multiple Kernel Learning. Journal of Machine Learning Research, 13(Aug):2465-2502.
M. Kloft, U. Brefeld, S. Sonnenburg, and A. Zien (2011). Lp-Norm Multiple Kernel Learning. Journal of Machine Learning Research, 12(Mar):953-997.
M. Kloft (2011). Lp-Norm Multiple Kernel Learning. Dissertation, Technische Universität Berlin.