Ruprecht-Karls-Universität Heidelberg
Bilder vom Neuenheimer Feld, Heidelberg und der Universität Heidelberg

Learning Similarity Metrics

Kursbeschreibung

Studiengang Modulkürzel Leistungs-
bewertung
BA-2010 AS-CL 8 LP
NBA AS-CL 8 LP
Master SS-CL, SS-TAC 8 LP
Magister - -
Dozenten/-innen Stefan Riezler, Artem Sokolov
Veranstaltungsart Hauptseminar
Erster Termin 23.04.2013
Zeit und Ort Di, 11:1512:45, INF 327 / SR 2 (SR)
Commitment-Frist 20.05.13.07.2013

Teilnahmevoraussetzungen

Foundations of Probability Theory und Statistics.

Leistungsnachweis

Regular and active participation, oral presentation.

Sprache

English

Inhalt

Similarity metrics are ubiquitous in machine learning and build the algorithmic foundation of numerous natural language processing applications, including information retrieval, machine translation, or paraphrasing. The goal of this seminar is to learn statistical similarity metrics from data instead of applying common similarity measures borrowed from, for example, linear algebra. For example, the emphasis is on the difference between applying a kernel and learning to adapt a kernel to the data; instead of applying cosine similarity to tfidf vectors we could rather learn the importance weights for tfidf entries, or replace standard singular value decomposition of term-document matrices with an optimal matrix factorization given a specific ranking problem.

Specific topics include:

  • Locality sensitive hashing
  • Distance metric learning
  • Kernel alignment

and others

Kursübersicht

Seminarplan

Datum Referent Material Thema
23.4. Artem Sokolov, Stefan Riezler Overview slides Introduction
30.4. Schigehiko Schamoni Bing Bai, Jason Weston, David Grangier, Ronan Collobert, Kunihiko Sadamasa, Yanjun Qi, Olivier Chapelle, and Kilian Weinberger. 2010. Learning to rank with (a lot of) word features. Information Retrieval 13, 3 Applications to NLP
7.5. Christoph Mayer Nello Cristianini, John Shawe-Taylor, André Elisseeff, Jaz S. Kandola: On kernel-target alignment. NIPS 2001: 367-373 Learning Kernels
and Features
14.5. (N.B. moved to 23.7.) Lyubov Nakryyko Koby Crammer, Joseph Keshet, Yoram Singer: Kernel Design using Boosting. NIPS 2002: 537-544
21.5. Michael Haas Brian Kulis & Kristen Grauman. Kernelized Locality-Sensitive Hashing. IEEE Trans. on Pattern Analysis and Machine Intelligence, vol. 34, no. 6, pp. 1092--1104, 2012 Learning to Hash
28.5. (N.B. moved to 25.6.) Angela Schneider Ruslan Salakhutdinov, Geoffrey Hinton. Semantic Hashing. International Journal of Approximate Reasoning 50 (2009) 969–978
4.6. Sariya Karimova Yair Weiss, Antonio Torralba, Robert Fergus. Spectral Hashing. NIPS, pp 1753-1760. 2008
11.6. cancelled Danny Rehl Killian Weinberger and Lawrence K. Saul, Distance Metric Learning for Large Margin Nearest Neighbor Classification, Journal of Machine Learning Research, 10(Feb):207--244, 2009 Metric Learning
18.6. Carolin Haas Prateek Jain, Brian Kulis, Inderjit Dhillon & Kristen Grauman. Online Metric Learning and Fast Similarity Search. NIPS, 2008
25.6. Angela Schneider Ruslan Salakhutdinov, Geoffrey Hinton. Semantic Hashing. International Journal of Approximate Reasoning 50 (2009) 969–978 Learning to Hash
2.7. Mareike Hartmann Chunhua Shen, Junae Kim, Lei Wang, Anton van den Hengel. "Positive Semidefinite Metric Learning Using Boosting-like Algorithms". Journal of Machine Learning Research, 13(Apr):1007−1036, 2012 Metric Learning
9.7. Bartosz Bogacz Maria-Florina Balcan, Avrim Blum, Santosh Vempala: Kernels as features: On kernels, margins, and low-dimensional mappings. Machine Learning 65(1): 79-94 (2006) Learning Kernels
and Features
16.7. Benjamin Heinzerling Zhuoran Wang and John Shawe-Taylor. 2010. A kernel regression framework for SMT. Machine Translation 24, 2 (June 2010), 87-102 Applications to NLP
23.7. Lyubov Nakryyko Koby Crammer, Joseph Keshet, Yoram Singer: Kernel Design using Boosting. NIPS 2002: 537-544 Learning Kernels and Features
zum Seitenanfang