Research

I am interested in various aspects of Machine Learning and Bayesian inference. It includes deterministic approximate methods and stochastic methods. Here you'll find preprints, papers, talks and notes.

Preprints

Low-Rank Factorization of Determinantal Point Processes for Recommendation (arXiv, pdf, AAAI pdf)
Mike Gartrell, Ulrich Paquet and Noam Koenigstein
The Fourteenth AAAI Conference on Artificial Intelligence, 2017
arXiv:1602.05436, 2016

Knowing What to Ask: A Bayesian Active Learning Approach to the Surveying Problem (pdf)
Yoad Lewenberg, Yoram Bachrach, Ulrich Paquet and Jeffrey S. Rosenschein
The Fourteenth AAAI Conference on Artificial Intelligence, 2017

Groove Radio: A Bayesian Hierarchical Model for Personalized Playlist Generation (pdf)
Shay Ben Elazar, Gal Lavee, Noam Koenigstein, Oren Barkan, Hilik Berezin, Ulrich Paquet and Tal Zaccai
WSDM, 2017

Publications

Sequential Neural Models with Stochastic Layers (arXiv, pdf, NIPS version, video) (NIPS 2016 oral paper)
Marco Fraccaro, Søren Kaae Sønderby, Ulrich Paquet and Ole Winther
Advances in Neural Information Processing Systems 29, 2016
arXiv:1605.07571, 2016

An Adaptive Resample-Move Algorithm for Estimating Normalizing Constants (arXiv, pdf)
Marco Fraccaro, Ulrich Paquet and Ole Winther
arXiv:1604.01972, 2016

An Efficient Implementation of Riemannian Manifold Hamiltonian Monte Carlo for Gaussian Process Models (pdf)
Ulrich Paquet and Marco Fraccaro
Supplementary material to arXiv:1604.01972, 2016

Bayesian Low-Rank Determinantal Point Processes (pdf)
Mike Gartrell, Ulrich Paquet and Noam Koenigstein
RecSys '16 Proceedings of the tenth ACM conference on Recommender Systems, 2016

Beyond Collaborative Filtering: The List Recommendation Problem (pdf)
Oren Sar Shalom, Noam Koenigstein, Ulrich Paquet and Hastagiri Vanchinathan
Proceedings of the 25th international conference on World Wide Web, 2016

Indexable Probabilistic Matrix Factorization for Maximum Inner Product Search (pdf)
Marco Fraccaro, Ulrich Paquet and Ole Winther
The Thirtieth AAAI Conference on Artificial Intelligence, 2016

Collective Noise Contrastive Estimation for Policy Transfer Learning (pdf, MSRC link, supplementary material, poster)
Weinan Zhang, Ulrich Paquet and Katja Hofmann
The Thirtieth AAAI Conference on Artificial Intelligence, 2016

Perturbation Theory for Variational Inference (pdf)
Manfred Opper, Marco Fraccaro, Ulrich Paquet, Alex Susemihl, and Ole Winther
NIPS 2015 Workshop on Advances in Approximate Bayesian Inferences, 2015

On the Convergence of Stochastic Variational Inference in Bayesian Networks (pdf)
Ulrich Paquet
NIPS 2014 Workshop on Advances in Variational Inference, 2014

A Scalable Bayesian Alternative to Density Estimation with a Bilinear Softmax Function (pdf)
Ulrich Paquet, Noam Koenigstein, and Ole Winther
NIPS 2014 Workshop on Personalization: Methods and Applications, 2014

Scalable Bayesian Modelling of Paired Symbols (arXiv, pdf)
Ulrich Paquet, Noam Koenigstein, and Ole Winther
arXiv:1409.2824, 2014

Speeding Up the Xbox Recommender System Using a Euclidian Transformation for Inner Product Spaces (pdf)
Yoram Bachrach, Yehuda Finkelstein, Ran Gilad-Bachrach, Liran Katzir, Noam Koenigstein, Nir Nice, and Ulrich Paquet
RecSys '14 Proceedings of the eighth ACM conference on Recommender Systems, 2014

A Large-scale Exploration of Group Viewing Patterns (pdf)
best paper runner-up
Allison J.B. Chaney, Mike Gartrell, Jake M. Hofman, John Guiver, Noam Koenigstein, Pushmeet Kohli, and Ulrich Paquet
TVX 2014: ACM International Conference on Interactive Experiences for Television and Online Video, 2014

Mining Large-scale TV Group Viewing Patterns for Group Recommendation (MSRC, pdf)
Allison Chaney, Mike Gartrell, Jake Hofman, John Guiver, Noam Koenigstein, Pushmeet Kohli, and Ulrich Paquet
Microsoft Technical Report, Number MSR-TR-2013-114, 2013

[Favourite paper]
Perturbative Corrections for Approximate Inference in Gaussian Latent Variable Models (arXiv, JMLR, pdf)
Manfred Opper, Ulrich Paquet, and Ole Winther
Journal of Machine Learning Research, Volume 14, 2857-2898, 2013

Xbox Movies Recommendations: Variational Bayes Matrix Factorization with Embedded Feature Selection (pdf)
Noam Koenigstein and Ulrich Paquet
RecSys '13 Proceedings of the seventh ACM conference on Recommender Systems, 2013

One-class Collaborative Filtering with Random Graphs: Annotated Version (arXiv, pdf)
Ulrich Paquet and Noam Koenigstein
arXiv:1309.6786, 2013
This arXiv version includes additional annotations to clarify some sections in the WWW 2013 paper.

One-class Collaborative Filtering with Random Graphs (pdf correcting eq. 9)
Ulrich Paquet and Noam Koenigstein
Proceedings of the 22nd international conference on World Wide Web, 999-1008, 2013
[Video] A preliminary talk, Large Scale Recommender System for The One Class Problem, from the Workshop on Large-scale Online Learning and Decision Making (Cumberland Lodge 2012), can be found here.

The Xbox Recommender System (pdf)
Noam Koenigstein, Nir Nice, Ulrich Paquet, and Nir Schleyen
RecSys '12 Proceedings of the sixth ACM conference on Recommender Systems, 281-284, 2012

Collaborative Learning of Preference Rankings (pdf)
Tim Salimans, Ulrich Paquet, and Thore Graepel
RecSys '12 Proceedings of the sixth ACM conference on Recommender Systems, 261-264, 2012

Transparent User Models for Personalization (pdf, appendix, talk)
Khalid El-Arini, Ulrich Paquet, Ralf Herbrich, Jurgen Van Gael and Blaise Agüera y Arcas
Proc. of 18th International Conference on Knowledge Discovery and Data Mining (KDD 2012), 2012

A Bayesian Treatment of Social Links in Recommender Systems (pdf)
Mike Gartrell, Ulrich Paquet, and Ralf Herbrich
Technical Report CU-CS-1092-12, University of Colorado, 2012

A Hierarchical Model for Ordinal Matrix Factorization (preprint pdf; final publication available at www.springerlink.com)
Ulrich Paquet, Blaise Thomson, and Ole Winther
Statistics and Computing, Volume 22(4), 945-957, 2012
Website and code: cogsys.imm.dtu.dk/ordinalmatrixfactorization; video (NIPS workshop 2007)

A penny for your thoughts? The value of information in recommendation systems (pdf)
Alexandre Passos, Jurgen Van Gael, Ralf Herbrich, and Ulrich Paquet
NIPS 2011 Workshop on Bayesian Optimization, Experimental Design, and Bandits, Sierra Nevada, Spain, 2011

Cumulant expansions for improved inference with EP in discrete Bayesian networks (pdf)
Manfred Opper, Ulrich Paquet, and Ole Winther
3rd NIPS Workshop on Discrete Optimization in Machine Learning, Sierra Nevada, Spain, 2011

Vuvuzelas & Active Learning for Online Classification (pdf)
Ulrich Paquet, Jurgen van Gael, David Stern, Gjergji Kasneci, Ralf Herbrich, and Thore Graepel
Computational Social Science and the Wisdom of Crowds Workshop (colocated with NIPS 2010), 2010

Large-scale Ordinal Collaborative Filtering (pdf)
Ulrich Paquet, Blaise Thomson, and Ole Winther
1st Workshop on Mining the Future Internet, Future Internet Symposium, Berlin, September 2010

Perturbation Corrections in Approximate Inference: Mixture Modelling Applications (JMLR, pdf)
Ulrich Paquet, Manfred Opper, and Ole Winther
Journal of Machine Learning Research, Volume 10, 935-976, 2009

Convexity and Bayesian Constrained Local Models (pdf)
Ulrich Paquet
CVPR (Computer Vision and Pattern Recognition) 2009
This paper contains ideas useful for face recognition. A summary poster is also available.

Improving on Expectation Propagation (pdf)
Manfred Opper, Ulrich Paquet, and Ole Winther
Advances in Neural Information Processing Systems 21, 1241-1248, 2009

Gaussian Process Modeling for Image Distortion Correction in EPI
Joseph W. Stevick, Sally G. Harding, Ulrich Paquet, Richard E. Ansorge, T. Adrian Carpenter, and Guy B. Williams
Magnetic Resonance in Medicine, Volume 59(3), 598-606, 2008

Bayesian Inference for Latent Variable Models (pdf)
Ulrich Paquet
PhD Thesis. Computer Laboratory, University of Cambridge, 2007

Gaussian Process Modeling for EPI Distortion Correction (pdf)
Joseph .W. Stevick, Sally G. Harding, Ulrich Paquet, Richard E. Ansorge, T. Adrian Carpenter, and Guy B. Williams
Joint Annual Meeting ISMRM-ESMRMB, 2007

Bayesian Hierarchical Ordinal Regression (pdf, ps)
Ulrich Paquet, Sean Holden, and Andrew Naish-Guzman
Proceedings of the International Conference on Artificial Neural Networks, 2005

On The Explicit Use Of Example Weights In The Construction Of Classifiers (pdf, ps)
Andrew Naish-Guzman, Sean Holden, and Ulrich Paquet
Proceedings of the International Conference on Artificial Neural Networks, 2005

Particle Swarms for Linearly Constrained Optimisation
Ulrich Paquet and Andries P. Engelbrecht
Fundamenta Informaticae, 76(1-2), 147-170, 2007

A New Particle Swarm Optimiser for Linearly Constrained Optimisation (pdf)
Ulrich Paquet and Andries P. Engelbrecht
Proceedings of the Congress on Evolutionary Computation, pages 227-233, 2003

Training Support Vector Machines with Particle Swarms (pdf)
Ulrich Paquet and Andries P. Engelbrecht
Proceedings of the International Joint Conference on Neural Networks, 2003

Training Support Vector Machines with Particle Swarms (pdf, ps)
Ulrich Paquet
MSc Thesis. Department of Computer Science, University of Pretoria, 2003

Talks and various notes

[Video] Large Scale Recommender System for The One Class Problem (video)
Ulrich Paquet, with Noam Koenigstein and the Xbox Recommendations team
Workshop on Large-scale Online Learning and Decision Making, 2012

[Video] Social recommender systems: a graphical model based approach (video)
Jurgen van Gael, with Stuart Elmore, Ralf Herbrich, Allen Jones, Ulrich Paquet, and David Stern.
PASCAL International Workshop on Social Web Mining, 2011
This describes Project Emporia, a fun news recommendation engine based on Twitter. This was our largest project at FUSE Labs at Microsoft Research Cambridge.

[Video] Large-scale Bayesian Inference for Collaborative Filtering (video, slides)
Ole Winther, with Ulrich Paquet, and Blaise Thomson
NIPS Workshop on Approximate Bayesian Inference in Continuous/Hybrid Models, 2007

[Video] Improving on Expectation Propagation (video, slides, poster)
Ulrich Paquet, with Ole Winther and Manfred Opper
NIPS Workshop on Approximate Bayesian Inference in Continuous/Hybrid Models, 2007

[Video] Perturbative Corrections to Expectation Consistent Approximate Inference (video, slides)
Manfred Opper, with Ulrich Paquet and Ole Winther
NIPS Workshop on Approximate Bayesian Inference in Continuous/Hybrid Models, 2007

Empirical Bayesian Change Point Detection (pdf)
Ulrich Paquet
2007