Center for Neural Computation Givat Ram Campus The Hebrew University Jerusalem 91904, Israel

Phone ++972-26585880 (W) ++972-25632890 (H) E-mail fink@cs.huji.ac.il

Michael Fink

Education

2004 - 2006 Ph.D. studies, Neural Computation, The Hebrew University Jerusalem, Israel. Thesis titled: “Feature Creation Processes in Multiclass Perceptual Learning” advised by Prof. Shimon Ullman (Weizmann Institute) and Prof. Gershon Ben-Shakhar (The Hebrew University)

2003 Graduate studies, California Institute of Technology, supervised by Prof. Pietro Perona, Pasadena, CA

2000 - 2002 M.A. Neural Computation, The Hebrew University, Jerusalem, Israel

1997 - 1999 B.Sc. Computer Science and Psychology, The Hebrew University, Jerusalem, Israel

Professional Experience

2006 summer Google Research, Developing methods for efficient large scale object recognition.

2005 summer Google Research, Initiated and tested the feasibility of mass-personalization. A project aimed at using ambient audio to effortlessly access personalized information, peer communities and targeted ads, while watching Television.

2001 ZUMO, Software engineer, developing an adaptive marketing platform based on automated learning of consumer preferences in peer to peer (P2P) networks.

2000 Adante Ltd., Algorithmic consultant, Pre-symptomatic diagnosis of Alzheimer using machine learning tools on immunological response profiles.

1999 fMRI Research Assistant to Prof. Dr. Axel Mecklinger at the Max Planck Institute for Neuropsychological Research, Leipzig, Germany

1998 - 1999 BioMediCom Ltd. Research project implementing Support Vector Machines for image segmentation enabling 3D ultrasound imagery

1997 - 1999 Electrophysiology Laboratory, The Hebrew University of Jerusalem, Programmer and Research Assistant

1993 - 1996 Military Intelligence, Israel Defense Force

Teaching Experience

2006-2007 Lecturer: Computation and Cognition

2004-2005 TA: Human Vision a Computational Approach (Prof. Yair Weiss)

2004-2005 TA: Computation and Cognition (Prof. Itamar Pitowsky)

2003 Mentor in the Caltech Freshman Summer Institute (FSI) project. Leading a research project aimed at real time multiclass object classification.

Grants, Awards and Scholarships

2003 - 2006 Horowitz Foundation scholarship for outstanding Ph.D. students

2002 - 2004 Young Scientist Research Grant from the Israeli Foundation Trustees

2000 - 2002 Interdisciplinary Center for Neural Computation full scholarship and stipend

1999 Max Planck Institute stipend

1998 Tutoring scholarship: mentoring deprived students

1997 Full tuition merit based scholarship

Selected Peer Reviewed Publications

Fink, M., Ben-Shakhar, G. and Ullman, S. Preferential Encoding of Features Distinctive for Multiple Categories (Under review)

Fink, M. and Ullman, S. From Aardvark to Zorro: A Benchmark of Mammal Images. Under review in the International Journal of Computer Vision (IJCV)

Amit, Y., Fink, M., Srebro, N. and Ullman, S. (2007) Uncovering Shared Structures in Multiclass Classification, The 24^th International Conference on Machine Learning (ICML) 2007

Fink, M. (2007) Online Learning of Search Heuristics, The 11^th International Conference on Artificial Intelligence and Statistics (AISTAT) 2007

Fink, M. (2007) Extracting shared representations in multiclass learning by trace-norm minimization. Invited speaker talk at the 12^th Winter Workshop on Computer Vision (WWCV) 2007

Fink, M., Covell, M., and Baluja, S. (2007) Mass personalization: social and interactive applications using sound-track identification. Journal of Multimedia Tools and Applications 2007

Fink, M., Shwartz-Shalev, S., Singer, Y., and Ullman, S. (2006) Online Multiclass Learning by Interclass Hypothesis Sharing. Proceedings of the International Conference on Machine Learning (ICML) 2006

Fink, M., Covell, M., and Baluja, S. (2006) Social and Interactive Television Applications Based on Real-Time Ambient Audio Identification. European Conference on Interactive Television (EuroITV) 2006 - best paper award.

Fink, M. and Ullman, S. (2005) Interclass Transfer as an Underlying Factor of Human Object Recognition, Neural Information Processing Systems workshop on Interclass Transfer (NIPS) 2005

Fink, M. (2004) Object classification from a single example utilizing class relevance Pseudo Metrics, Proceedings of the Neural Information Processing Systems (NIPS), 2004

Levi, K., Fink, M. and Weiss, Y. (2004). Learning From a Small Number of Training Examples by Exploiting Object Categories, Workshop of Learning in Computer Vision (CVPR) 2004

Fink, M. and Perona, P. (2003). Mutual Boosting for Contextual Inference. Proceedings of the Neural Information Processing Systems (NIPS), 2003

Fink, M., Ben-Shakhar, G. and Horn, D. (2003). Empirical Evidence and Theoretical Analysis of Feature Creation during Category Acquisition. Connectionist Models of Cognition and Perception II, Canterbury, UK 2003

Patents

GP-701-00-US: Detecting Repeating Content in Broadcast Media

GP-636-00-US: Social and Interactive Applications for Mass Media

Miscellanea

Organized the NIPS’05 workshop titled: “Interclass transfer: why learning to recognize many classes might be easier than learning to recognize just one”, with Andras Ferencz and Shimon Ullman

Languages: C++, Perl, Matlab

Platforms: Unix, Windows