Université de Montréal

Faculté des arts et des sciences -Secteur des sciences
Département d'informatique et de recherche opérationnelle


Département d'informatique et de recherche opérationnelle
Fr   En

McGill - UdeM - MITACS Machine Learning Seminars


(PNG)

Welcome to the official web page of the McGill-UdeM-MITACS Machine Learning Seminars. This initiative from University of Montreal and McGill University aims at

  • diffusing machine learning research done at both universities and stimulating discussion and collaboration
  • offering introductive talks on machine learning subjects
  • hosting talks by invited speakers

The seminars are held successively at University of Montreal and McGill University, and delivered in English. The time and location are specified for each seminar.

For questions or suggestions, contact:

  • Dumitru Erhan (erhandum[at]iro.umontreal.ca), PhD student at UdeM (LISA lab, DIRO),
  • Doina Precup (dprecup[at]cs.mcgill.ca), Assistant Professor at McGill,

Also, a mailing list in computational neuroscience for researchers in Quebec has just been setup by François Rivest, student at the Département d’Informatique et de Recherche Opérationnelle. The objective of this initiative is to stimulate research in computational neuroscience, but more importantly, to help develop strong multi-disciplinary collaborations in that field in Quebec.

To register on the list, go to http://www.listes.umontreal.ca/wws/info/neuralcomp. To publish on the list, e-mail to neuralcomp[at]listes.umontreal.ca. Do not hesitate to forward events of interest for computational neuroscience researchers on this list.


-  Next Seminar

Unlocking Brain-Inspired Computer Vision: a Multi-Disciplinary, High-Throughput Approach
by Nicolas Pinto
Department of Brain and Cognitive Sciences
Massachusetts Institute of Technology
Friday, November 20 at 14:30, AA-3195 (UdeM)

The construction of artificial vision systems and the study of biological vision are naturally intertwined as they represent simultaneous efforts to forward and reverse engineer systems with similar goals. While exploration of the neuronal substrates of visual processing provides clues and inspiration for artificial systems, artificial systems can in turn serve as important generators of new ideas and working hypotheses. However, while systems neuroscience has so far provided inspiration for some of the "broad-stroke" properties of the visual system (e.g. hierarchical organization, synaptic integration of inputs and threshold, normalization, plasticity, etc), much is still unknown. Even for those qualitative properties that most biological-inspired models hold in common, experimental data currently provide little constraint on their key parameters. Consequently, it is difficult to truly evaluate a set of computational ideas, since the performance of any one model depends strongly on its particular instantiation - e.g. the size of the pooling kernels, the number of units per layer, exponents in normalization operations, etc. Since the number of such parameters (explicit or implicit) is very large, and the typical computational cost of evaluating one particular model is high, the space of possible model instantiations usually goes largely unexplored. Compounding the problem, even if a set of computational ideas are on the right track, the instantiated "scale" of those ideas is typically small (e.g. in terms of dimensionality and amount of learning experience provided). Thus, when a model fails to approach the abilities of the visual system, we are left uncertain whether this failure is because we are missing a fundamental idea, or because the correct "parts" have not been tuned correctly, assembled at sufficient scale, or provided with sufficient natural experience.

To pave a possible way forward, we have begun developing a high-throughput approach to expansively explore a large range of biologically-inspired models - including models of larger, more realistic scale - leveraging recent advances in commodity stream processing hardware (high-end GPUs and Playstation 3’s Cell processors) and scientific cloud computing (e.g. Amazon EC2). In analogy to high-throughput screening approaches in molecular biology and genetics, we generated and trained thousands of potential network architectures and parameter instantiations, and "screened" the visual representations produced by these models using an object recognition task. From these candidate models, the most promising were selected for further analysis. We have shown that this approach can yield significant, reproducible gains in performance across an array of basic object recognition tasks, consistently outperforming a variety of state-of-the-art purpose-built vision systems from the literature, and that it can offer insight into which computational ideas are most important for achieving this performance.

As the scale of available computational power continues to expand, we believe that this approach holds great potential both for accelerating progress in artificial vision, and for generating new, experimentally-testable hypotheses for the study of biological vision.


-  Previous seminars

Here is the list of past seminars, with their corresponding slides::

Seminars 2008-2009

[18/09/2009] Learning from Multiple Partially Observed Views, by Massih-Reza Amini [abstract] [pdf]
[16/07/2009] Probability, Pattern Recognition and Optimization Techniques for Interactive Media, by Christopher Pal [abstract]
[22/06/2009] Reinforcement learning and apprenticeship learning for robotic control, by Andrew Ng [abstract]
[27/03/2009] Machine Learning and Computational Linguistics: Tools of the Trade (or, The Data Acquisition Mental Bottleneck), by Jeremy Barnes [abstract]
[13/03/2009] Recent developments in learning deep networks, by Geoffrey Hinton [abstract]
[30/01/2009] New architectures and algorithms for solving deep-memory POMDPs, by Daan Wierstra [abstract]
[21/11/2008] Analysis of EEG data by means of ordinal pattern distributions, by Mathieu Sinn [abstract]
[17/11/2008] Cheap and Fast — But is it Good?
Evaluating Nonexpert Annotations for Machine Learning Tasks
, by Rion Snow
[abstract]
[11/11/2008] Using fast weights to improve Persistent Contrastive Divergence, by Tijmen Tieleman [abstract]
[16/10/2008] Visualizing high-dimensional data using t-SNE, by Geoffrey Hinton [abstract]
[10/10/2008] Large Scale Learning for Natural Language Processing, by Ronan Collobert [abstract]
[22/08/2008] Differentiable Sparse Coding, by J. Andrew Bagnell [abstract]

Seminars 2007-2008

[30/06/2008] Hierarchical learning and decision making by Stuart Russell [abstract]
[26/06/2008] Artificial Intelligence in Seven Years? by Léon Bottou [abstract]
[20/06/2008] Learning hierarchical representations of natural images by Mike Lewicki [abstract] [pdf]
[27/05/2008] Learning in the Sample Compression Framework by Mohak Shah [abstract] [pdf]
[13/05/2008] Deep belief networks are universal approximators and a recurrent neural network that learns to remember par Ilya Sutskever [abstract] [pdf] [code]
[30/04/2008] Learning Deep Hierarchies of Sparse and Invariant Features by Yann LeCun [abstract]
[18/04/2008] Restricted Boltzmann machines: Performance and behavior on image databases and future plans by Karol Gregor [abstract]
[08/04/2008] Aggregate Markov Decision Processes by Hasan Mirza [abstract]
[25/03/2008] Image Classification with Higher-Order Neural Models by James Bergstra [abstract] [pdf]
[25/03/2008] Deep Learning with Denoising Autoencoders by Pascal Vincent [abstract] [pdf]
[14/03/2008] Discriminative Methods with Structure by Simon Julien-Lacoste [abstract] [ppt]
[04/03/2008] A Stochastic Algorithm for Partially Observable Markov Decision Processes (POMDPs) by François Laviolette [abstract]
[19/02/2008] Modelling Image Patches With A Directed Hierarchy Of Markov Random Fields by Simon Osindero [abstract] [pdf]
[22/01/2008] Open Problems in Statistical Machine Translation by Roland Kuhn [abstract] [ppt]
[30/11/2007] A Bayesian approach to analyze water tank signals in the Pierre Auger experiment by Balazs Kegl [abstract] [pdf]
[16/11/2007] Bayesian Reinforcement Learning by Mohammad Ghavamzadeh [abstract]
[02/11/2007] Preliminary steps toward a probabilistic, decision theoretic model of dynamic scene understanding by Nando de Freitas [abstract]
[19/10/2007] Learning the 2-D Topology of Images by Pascal Lamblin [abstract] [pdf]
[28/09/2007] How Many Clusters? An Information-Theoretic Perspective by Susanna Still [abstract]
[14/09/2007] Optimal Causal Inference by Susanna Still [abstract] [pdf]

Seminars 2006-2007

[28/09/2006] Modelling high-dimensional sequential data using distributed hidden state by Geoffrey Hinton [abstract]
[05/10/2006] Program Verification using Reinforcement Learning by Sami Zhioua [abstract]
[24/10/2006] An Upper Bound on the Convergence Time of the Gibbs Sampler in Ising Models by Yuichi Shiraishi [abstract] [ps]
[31/10/2006] Dirichlet processes: interpretations, inference and extensions by Aaron Courville [abstract] [pdf]
[07/11/2006] Bayesian Ranking using Factor Graphs and Expectation Propagation by Dumitru Erhan [abstract] [ppt]
[14/11/2006] PAC-learning of Markov models with hidden state by Doina Precup [abstract] [pdf]
[21/11/2006] Greedy Layer-Wise Training of Deep Networks by Yoshua Bengio [abstract] [ps/nips] [pdf/nips]
[29/11/2006] Decision Making under Parameter Uncertainty by Shie Mannor [abstract] [pdf]
[12/12/2006] Segment Class Models for Music by Darrell Conklin [abstract]
[14/12/2006] Towards Simple and Effective Connectionist Nonparametric Estimation of Probability Density Functions by Edmondo Trentin [abstract] [pdf]
[10/01/2007] Modeling Appearance Patterns in Image Sets by Matt Toews [abstract] [pdf]
[17/01/2007] Hierarchical Boosting and Filter Generation by Marc-Olivier LaBarre [abstract] [pdf]
[25/01/2007] Wake-sleep Algorithm for Representational Learning by Hamid Reza Maei [abstract] [ppt]
[28/02/2007] Machine Learning Based Robotics by Greg Grudic [abstract] [pdf]
[14/03/2007] Topmoumoute Online Natural Gradient Algorithm by Nicolas Le Roux [abstract]
[21/03/2007] An Anytime Error Minimization Search for Online Policy Improvement in Large POMDPs by Stephane Ross [abstract] [pdf]
[28/03/2007] Real Neurons for Machine Learning by François Rivest [abstract] [flash]
[04/04/2007] Sampling-based Planning with Approximate and Learned MDP Models par Cosmin Paduraru [abstract] [pdf]
[11/04/2007] Majority Votes by François Laviolette [abstract] [pdf]
[18/04/2007] Sequence kernels for Speaker Verification using Support Vector Machines by Jérôme Louradour [abstract] [pdf]
[25/04/2007] Constituent Parsing by Classification by Joseph Turian [abstract]
[02/05/2007] Optimizing Disease Outbreak Detection Methods Using Reinforcement Learning by Masoumeh Izadi [abstract] [ppt]
[09/05/2007] Towards the construction of spatio-temporal data patterns in a road trafic urban network by Marc Joliveau [abstract] [pdf]
[16/05/2007] Regret to the Best vs. Regret to the Average by Eyal Even-dar [abstract]
[30/05/2007] Automated Hierarchy Discovery for Planning in Partially Observable Environments by Laurent Charlin [abstract] [pdf]
[13/06/2007] An Empirical Evaluation of Deep Architectures on Problems with Many Factors of Variation by Hugo Larochelle [abstract] [pdf]

Best Of NIPS/IJCAI 2007

  1. iLSTD: Eligibility Traces and Convergence Analysis [pdf], slides [pdf]
    from Alborz Geramifard, Michael Bowling, Martin Zinkevich, Richard Sutton,
    presented by Cosmin Paduraru

  2. Direct Code Access in Self-Organizing Neural Networks for Reinforcement Learning [pdf], slides [pdf]
    from Ah-Hwee Tan,
    presented by Marc G. Bellemare

  3. Efficient Failure Detection On Mobile Robots Using Particle Filters With Guassian Process Proposals
    from Christian Plagemann, Dieter Fox, Wolfram Burgard,
    presented by Amin Atrash

  4. Solving POMDPs Using Quadratically Constrained Linear Programs [pdf], slides [pdf]
    from Christopher Amato, Daniel S. Bernstein, Shlomo Zilberstein,
    presented by Robert Kaplow

  5. Bayesian Inverse Reinforcement Learning [pdf], slides [pdf]
    from Deepak Ramachandran, Eyal Amir,
    presented by Monica Dinculescu

  6. Linearly-solvable Markov decision problems [pdf], slides [pdf]
    from Emanuel Todorov,
    presented by Jordan Frank

  7. An Efficient Method for Gradient-Based Adaptation of Hyperparameters in SVM Models [pdf], slides [pdf]
    from Sathiya Keerthi, Vikas Sindhwani, Olivier Chapelle,
    presented by Dumitru Erhan

  8. Efficient Learning of Sparse Representations with an Energy-Based Model [pdf], slides [pdf]
    from MarcAurelio Ranzato, Christopher Poultney, Sumit Chopra, Yann LeCun,
    presented by Pascal Lamblin

  9. Multi-Task Feature Learning [pdf], slides [pdf]
    from Andreas Argyriou, Theos Evgeniou, Massimiliano Pontil,
    presented by Olivier Delalleau

  10. Active learning for misspecified generalized linear models [pdf], slides [pdf]
    from Francis Bach,
    presented by Yoshua Bengio

  11. Analysis of Representations for Domain Adaptation [pdf], slides [pdf]
    from Shai Ben-David, John Blitzer, Koby Crammer, Fernando Pereira,
    presented by Marina Sokolova

  12. No-regret algorithms for Online Convex Programs [pdf], slides [ppt]
    from Geoffrey Gordon,
    presented by Nicolas Chapados

Seminars 2005-2006

[07/09/2005] A Hybrid Pareto Model for Conditional Density Estimation of Asymmetric Fat-Tailed Data by Julie Carreau [abstract] [pdf]
[14/09/2005] A Brief Tutorial on Reinforcement Learning par Doina Precup [abstract] [pdf]
[21/09/2005] The K-Shortest-Paths Approach to Approximate Dynamic Programming by Nicolas Chapados [abstract] [pdf]
[28/09/2005] Active Learning in Partially Observable Markov Decision Processes by Robin Jaulmes [abstract] [ppt]
[05/10/2005] Music/Voice and Genre Classification by Norman Casagrande and James Bergstra [abstract] [pdf]
[12/10/2005] Musical Rhythm Similarity Measures with Symbolic Input: Models, Algorithms and Applications by Godfried Toussaint [abstract]
[19/10/2005] Neural Networks in Graphical Domains by Marco Gori [abstract] [pdf] [key]
[26/10/2005] Automatic learning of adaptive treatment strategies using kernel-based reinforcement learning by Joelle Pineau [abstract] [pdf]
[02/11/2005] Echo State Networks: a Tutorial by Dan Popovici [abstract] [ppt]
[09/11/2005] Dimensionality reduction for policy evaluation in MDPs by Philipp Keller [abstract] [pdf]
[16/11/2005] Convolutional Neural Networks and Vision Applications by Yoshua Bengio [abstract] [pdf]
[16/11/2005] RTBSS: An Online POMDP Algorithm for Complex Environments by Sébastien Pacquet [abstract] [pdf]
[30/11/2005] Empirical Artificial Intelligence by Rich Sutton [abstract] [pdf]
[10/01/2006] Onset Detection with Artificial Neural Networks for MIREX2005 by Alexandre Lacoste [abstract] [ppt]
[17/01/2006] Predictive State Representations by Masoumeh Tabaeh Izadi [abstract] [ppt]
[24/01/2006] Highlights of Hinton’s Contrastive Divergence Pre-NIPS Workshop by Yoshua Bengio and Pascal Lamblin [abstract] [ppt]
[07/02/2006] La Gestion Active de Portefeuille by Marc Boucher [abstract] [ppt]
[28/02/2006] Junction Tree Harmonisation by Jean-François Paiement [abstract] [pdf]
[20/03/2006] Effective learning of deep belief nets, with application to handwritten digits and face images by Simon Osindero [abstract]
[21/03/2006] Learning to Control an Octopus Arm with Gaussian Process Temporal Difference Learning by Yaakov Engel [abstract] [pdf]
[22/03/2006] Reinforcement Learning with Gaussian Processes by Yaakov Engel [pdf]
[24/03/2006] Performance guarantees and stability of approximate ERM by Alexander Rakhlin [abstract]
[27/03/2006] Algorithmes d’inspiration géométrique, corrigés pour l’apprentissage en haute dimension by Pascal Vincent [abstract]
[11/04/2006] Discriminant Mixture of 3D Molecular Surface Models by Pascal Lamblin [abstract] [pdf]
[18/04/2006] Continuous Neural Network by Nicolas Le Roux [abstract] [pdf]
[25/04/2006] Detection and Control for the Treatment of Epilepsy by Robert D. Vincent [abstract] [pdf]
[28/04/2006] Regret minimization under partial monitor by Gilles Stoltz [abstract]
[02/05/2006] Collaborative Filtering for Drug Discovery by Dumitru Erhan [abstract] [pdf]
[09/05/2006] Observation Space Reduction and Active Learning for Partially Observable Markov Decision Processes by Amin Atrash [abstract]
[17/05/2006] Music Plus One by Christopher Raphael [abstract]
[18/05/2006] Decision-making with Predictive State Representations (PSRs) by Michal James [abstract]
[13/06/2006] Neural Basis of Learning by François Rivest [abstract] [pdf]
[20/06/2006] The decoder in statistical machine translation: how does it work? by Alexandre Patry [abstract] [pdf]
[27/06/2006] Global Optimization and Evolutionary Algorithms by Javad Sadri [abstract]

Best-of NIPS 2006

  1. Temporal Abstraction in Temporal-difference Networks [pdf], slides [pdf]
    from Richard Sutton, Eddie Rafols, Anna Koop,
    presented by Marc Gendron-Bellemare

  2. Learning Influence among Interacting Markov Chains [pdf], slides [pdf]
    from Dong Zhang, Daniel Gatica-Perez, Samy Bengio, Deb Roy,
    presented by Pablo Samuel Castro

  3. Robust Fisher Discriminant Analysis [pdf], slides [ppt]
    from Seung-Jean Kim, Alessandro Magnani, Stephen Boyd,
    presented by Erick Delage

  4. TD(0) Leads to Better Policies than Approximate Value Iteration [pdf], slides [pdf]
    from Benjamin Van Roy,
    presented by Philipp Keller
  5. Metric Learning by Collapsing Classes [pdf], slides [ppt]
    from Amir Globerson, Sam Roweis,
    presented by Dumitru Erhan

  6. Towards Human-level AI: Yann LeCun’s talk, slides [pdf]
    from Yann LeCun,
    presented by Alexandre Lacoste

  7. Highlights on the ’Inductive Transfer : 10 Years Later’ Workshop [html], slides [pdf]
    from Danny Silver, Goekhan Bakir, Kristin Bennett, Rich Caruana, Massimiliano Pontil, Stuart Russell, Prasad Tadepalli,
    presented by Hugo Larochelle
  8. Rodeo: Sparse Nonparametric Regression in High Dimensions [pdf], slides [pdf]
    from John Lafferty, Wasserman Larry,
    presented by Olivier Delalleau
  9. Beyond Gaussian Processes: On the Distributions of Infinite Networks [pdf], slides [pdf]
    from Ricky Der, Daniel Lee,
    presented by Julie Carreau
  10. Inferring Motor Programs From Images of Handwritten Digits [pdf], slides [pdf]
    from Geoff Hinton, Vinod Nair,
    presented by Nicolas Le Roux
  11. Gaussian Process Dynamical Models [pdf], slides [ppt]
    from Jack Wang, Fleet David, Hertzmann Aaron,
    presented by Nicolas Chapados
  12. Divergences, surrogate loss functions and experimental design [pdf], slides [pdf]
    from XuanLong Nguyen, Wainwright Martin, Jordan Michael,
    presented by James Bergstra
  13. Bayesian Sparse Sampling for On-line Reward Optimization [pdf], slides [ppt]
    from Dale Schuurmans, au "Workshop on Value of Information in Inference, Learning and Decision-Making",
    presented by Dan Popovici
  14. Distance Metric Learning for Large Margin Nearest Neighbor Classification [pdf], slides [pdf]
    from Kilian Q. Weinberger, John Blitzer, Lawrence K. Saul,
    presented by Gilles Godbout
  15. Infinite latent feature models and the Indian buffet process [pdf]
    from Tom Griffiths, Ghahramani Zoubin,
    presented by Aaron Courville

-  Former Web Page

To access the former LISA seminars web page, click here