- Lisa_teatalk - mailman.iro.umontreal.ca

[Tea Talk] Bradly Stadie (UC Berkeley) Wed Feb 21 3:00PM AA1360
by Michael Noukhovitch 22 Fév '18

22 Fév '18

Sorry for the lateness, please note this talk is a normal tea talk and it is *tomorrow!* If you want to meet with the speaker, please email me and I'll set it up! This week we have *Bradly Stadie* from *UC Berkeley* giving a talk on *Wednesday Feb 21* at *3:00PM* in room *AA1360*. See you there! Michael *TITLE* Towards Quickly Learning Agents *KEYWORDS *Meta-learning, Reinforcement Learning, Imitation Learning *ABSTRACT*We consider the problems of meta reinforcement learning, wherein the goal is to train agents that can learn new tasks quickly. We suggest 5 major sub-problems that are crucial for enabling quickly learning agents: Exploration, Imitation, Hierarchy/Planning, Credit Assignment, and Control. The talk will examine the sub-problems of exploration and imitation and their interplay with quickly learning agents in more depth. In particular, we examine: 1) Third Person Imitation: which enables agents to imitate other agents that differ significantly from themselves. 2) One Shot Imitation: which enables a robot to see a single video of a demonstration and copy the demonstrated behavior. 3) E-MAML: a meta-RL algorithm that achieves significantly better exploration than previous meta-RL algorithms. *BIO* Bradly Stadie is a PhD student at UC Berkeley in Pieter Abbeel’s lab. Previously, he spent 2 years working at Open AI under Ilya Sutskever. Bradly’s work has focused on the fields of exploration, imitation, and meta learning with applications to games, robotics, and causal inference. Prior to his interest in AI, Bradly studied mathematics at the University of Chicago, where his work focused on representation theory of locally compact groups and stochastic differential equations.

2 2

[DIRO Talk] Pieter-Jan Kindermans (Google/Ghent) Thur Feb 22 11:00AM AA1355
by Michael Noukhovitch 21 Fév '18

21 Fév '18

Continuing our special DIRO talk week, we have *Pieter-Jan Kindermans* from *Ghent University and Google Brain* giving a talk on *Thursday Feb 22* at *11:00AM* in room *AA1355*. As always, email me if you'd be interested in speaking with him! This talk should be good because I've heard the speaker is great at explaining ML! Michael *TITLE *Towards a reliably understanding and visualization of deep neural networks *KEYWORDS *Explainable ML, Testing Deep Learning, Deep Learning Theory *ABSTRACT*Deep learning has transformed the field of machine learning. Empirically these methods work brilliantly but it is difficult to understand what exactly they have learned. If this were possible, we would have a perfect understanding of our networks. The community has spend quite a lot of effort on visualising what a neural network learns. However while many methods were invented to better understand deep neural networks, we can show that these methods do not produce the theoretically correct explanation for a linear model, which is a simple neural network. Despite this they are used on multi-layer networks with millions of parameters. We present the idea of creating unit tests for explanations. The idea behind the unit test is that while it might be impossible to define what a good explanation is, it is much easier to detect failure cases. Hence, by creating more and more reliable unit tests, the community can refine its methods iteratively and converge to a good solution. Based on an analysis of linear models we propose a generalization that yields two explanation techniques (PatternNet and PatternAttribution) that are theoretically sound for linear models, pass our unit test and produce improved explanations for deep networks. *BIO*Pieter-Jan Kindermans obtained his PhD degree from Ghent Unviersity in 2014. From 2014 to 2017 he was a postdoc, as a Marie-Curie fellow, in the lab of Klaus-Robert Müller in Berlin. Currently he is a Brain resident at Google. Initially he worked on unsupervised learning for Brain-Computer Interfaces (BCI) and demonstrated that unsupervised learning can replace traditional supervised classifiers. A further extension of this work, in collaboration with the universities of Freiburg and Ghent, ensured that the unsupervised decoder is guarantueed to converge to the optimal supervised solution. This project was nominated for the BCI award in 2017.In addition to his work in BCI, he also explored different subfields of machine learning, including deep learning, large scale training and its applications such as molecular chemistry. His current focus is on better understanding the operation of a deep neural network. His ultimate goal is to make deep learning a knowledge extraction tool for science.

1 0

[DIRO Talk] Marc Law (U of T) Tue Feb 20 11:00AM PCM Z240
by Michael Noukhovitch 20 Fév '18

20 Fév '18

We will have three special DIRO talks next week and kicking it off is *Marc Law*, from University of Toronto giving a talk on *Tuesday Feb 20* at* 11AM* in room *PCM Z240*. *Note that this is on *Tuesday*, and in a different room* Checks all the boxes for a fantastic talk, but you should come see for yourself! Michael *TITLE *Shallow and Deep Metrics for Machine Learning and Computer Vision *KEYWORDS* computer vision, few/zero-shot learning, metric learning *ABSTRACT* Similarity functions and distance metrics are used in many machine learning and computer vision contexts such as clustering, k-nearest neighbors classification, support vector machine, information/image retrieval, visualization etc. Traditionally, machine learning methods fixed sample representations and the used metric before learning a model optimized for the target task. Metric learning approaches, which learn the employed metric in a supervised way, have been proposed to increase performance on tasks such as clustering. In particular, they have shown great generalization performance to compare objects from categories that were not seen during training (for instance in face verification or few-shot learning). In this talk, I will talk about different shallow and deep metric learning approaches optimized for clustering and reducing model complexity. In the clustering task, I will present efficient approaches to learn a metric in a supervised or weakly supervised way. In the model complexity context, I will present approaches to limit the rank of shallow approaches, or reduce the dimensionality of a pretrained deep neural network to perform visualization or increase zero-shot learning performance. *BIO* Marc Law received a PhD in Computer Science from Université Pierre et Marie Curie (Paris, France) in 2015. He was a visiting research scholar in the team of Professor Eric Xing at the School of Computer Science, Carnegie Mellon University in 2015~2016. Since June 2016, he is a postdoctoral fellow in the Department of Computer Science (Machine Learning group) at the University of Toronto under the supervision of Professor Raquel Urtasun and Professor Richard Zemel. During his PhD, he worked mostly on distance metric learning applied to different contexts of computer vision and web archiving. He is currently working on deep learning. His main focus is to propose scalable machine learning methods. He received an award for best PhD from the French Association for Artificial Intelligence in 2016.

2 1

[Tea Talk] Jacob Steinhardt (Stanford) Fri Feb 16 10:30AM AA1360
by Michael Noukhovitch 16 Fév '18

16 Fév '18

This week we have *Jacob Steinhardt* from Stanford giving a special DIRO talk on *Fri Feb 16* at* 10:30AM* in room *AA1360*. Don't be fooled, this talk should be the real deal! Michael *TITLE *Provably Secure Machine Learning *KEYWORDS *Security, Optimization, AI Safety, Adversarial *ABSTRACT* The widespread use of machine learning systems creates a new class of computer security vulnerabilities where, rather than attacking the integrity of the software itself, malicious actors exploit the statistical nature of the learning algorithms. For instance, attackers can add fake data (e.g. by creating fake user accounts), or strategically manipulate inputs to the system once it is deployed. So far, attempts to defend against these attacks have focused on empirical performance against known sets of attacks. I will argue that this is a fundamentally inadequate paradigm for achieving meaningful security guarantees. Instead, we need algorithms that are provably secure by design, in line with best practices for traditional computer security. To achieve this goal, we take inspiration from robust statistics and robust optimization, but with an eye towards the security requirements of modern machine learning systems. Motivated by the trend towards models with thousands or millions of features, we investigate the robustness of learning algorithms in high dimensions. We show that most algorithms are brittle to even small fractions of adversarial data, and then develop new algorithms that are provably robust. Additionally, to accommodate the increasing use of deep learning, we develop an algorithm for certifiably robust optimization of non-convex models such as neural networks. *BIO*Jacob Steinhardt is a graduate student in artificial intelligence at Stanford University working with Percy Liang. His main research interest is in designing machine learning algorithms with the reliability properties of good software. So far this has led to the study of provably secure machine learning systems, as well as the design of learning algorithms that can detect their own failures and generalize predictably in new situations. Outside of research, Jacob is a technical advisor to the Open Philanthropy Project, and mentors gifted high school students through the USACO and SPARC summer programs.

1 1

No [Tea Talk] this week
by Michael Noukhovitch 06 Fév '18

06 Fév '18

Good luck on ICML submissions!

1 0

Fwd: Articles sur l'Intelligence artificielle dans le "O Mile-Ex"
by Yoshua Bengio 03 Fév '18

03 Fév '18

---------- Message transféré ---------- De : Morin Danielle <danielle.morin(a)umontreal.ca> Date : 2 février 2018 à 12:43 Objet : RE: Articles sur l'Intelligence artificielle dans le "O Mile-Ex" À : Myriam Côté <myriam.cote(a)rd.mila.quebec> Cc : Bengio Yoshua <yoshua.bengio(a)umontreal.ca>, Myriam CÔTÉ < cotemyri(a)gmail.com> ------------------------------ Good Morning, The piece of news concerning the installation of the hub in artificial intelligence in the building of O Mile-Ex situated on the street St-Urbain was made public this morning by Radio-Canada and in LaPresse +. The information concerning the various tenants of the building, including the MILA, being made public, thus the owner confirmed the "hub" to the journalists. This piece of news(short story) precedes the official announcement and we wanted you are informed about it. Furthermore, the owner of the building will grant an interview this evening to Téléjournal of 6:00 pm of Radio-Canada and to RDI economy. Best regards, *Danielle* *Morin* Directrice générale par intérim MILA – Institut québécois d’intelligence artificielle Téléphone : 514-343-2186 <(514)%20343-2186> Cellulaire : 514-261-0597 <(514)%20261-0597> Danielle.morin(a)umontreal.ca Here are the links towards the Web article of Radio-Canada and that of the exercise book Business(Affairs) of LaPresse +. http://ici.radio-canada.ca/nouvelle/1081709/technologie- montreal-saint-urbain-intelligence-artificielle http://plus.lapresse.ca/screens/47a384ca-5655-4d99- 92d7-ebc482759a00__7C___0.html?utm_medium=Email&utm_ campaign=Microsite+Share&utm_content=Screen

1 0

Fwd: telluride workshop
by Yoshua Bengio 02 Fév '18

02 Fév '18

---------- Forwarded message ---------- From: Tobi Delbruck (INI) <tobi(a)ini.uzh.ch> Date: 2018-02-02 2:28 GMT-05:00 Subject: telluride workshop To: Yoshua Bengio <yoshua.bengio(a)umontreal.ca> Hi Yoshua, if you or someone on your team could share this year's telluride workshop to g+ or other ML community or related forums it would be great. (I don't know these communities very well). The webpage is now up and topic areas have been selected and the workshop is now open for participation. https://sites.google.com/view/telluride2018/home It should be quite fun this year; would be great if people from Montreal could participate. Best, Tobi -- Prof. Tobi Delbruck, Inst. for Neuroinformatics, UZH-ETH Zurich Sensors Group: http://sensors.ini.uzh.ch Personal page: http://www.ini.unizh.ch/~tobi Building 55, Room G84, Winterthurerstr. 190, CH-8057 Zurich, Switzerland <https://maps.google.com/?q=Winterthurerstr.+190,+CH-8057+Zurich,+Switzerlan…> Tel lab +41 44 635 3038, cell +41 76 629 1500

1 0

[Tea Talk] William Hamilton (Stanford) Fri Feb 2, 10:30AM, AA1360
by Michael Noukhovitch 01 Fév '18

01 Fév '18

*Note:* This is the same as the DIRO colloqium email, I'm just making sure that everyone on the mailing lists gets it! Sorry for the late email! This week we have a PhD Candidate from Stanford, *William Hamilton,* giving a talk on *Friday Feb 2* at* 10:30AM* in room *AA1360*. Make sure to chart your way to this tea talk and invite all your friends! Michael *KEYWORDS *ML on Graphs, Massive Structured Datasets, Social Science *TITLE* Machine Learning for Computational Social Science *ABSTRACT* The combination of machine learning and massive social datasets has the potential to revolutionize our ability to predict and understand human behavior. However, machine learning on large social datasets is difficult because these datasets tend to be noisy, dynamic, and involve graph-structured relationships (e.g., social networks between users)—while traditional machine learning tools are largely designed for static datasets comprised of simple Euclidean vectors or grids. In this talk, I will describe new methods that I have developed for machine learning on massive, graph-structured social datasets. The technical focus of the talk will be on techniques for graph embedding, i.e., representation learning on graph-structured data. In the first part, I will describe how I have used graph embedding techniques to enable diverse social applications—from modeling cultural change to predicting conflict between online communities. In the second part, I will describe a new graph embedding framework, called GraphSAGE, that can scale to datasets that are orders of magnitude larger than previous approaches and that has been deployed at the website Pinterest to power a recommender system serving over 200 million users. I will close the talk with a general outlook on social AI technologies, including future technical directions and important ethical considerations. *BIO* William (Will) Hamilton is a PhD Candidate in Computer Science at Stanford University, working jointly in the NLP and SNAP groups. He is co-advised by Dan Jurafsky and Jure Leskovec, and his interests lie at the intersection of machine learning, network science, natural language processing, and computational social science. Will's research is supported by the SAP Stanford Graduate Fellowship and an NSERC PGS-D Grant. Prior to coming to Stanford, Will completed a BSc and MSc at McGill University, where he studied machine learning under the supervision of Joelle Pineau and was the 2014 recipient of the Canadian AI Master's Thesis Award.

1 0

Fwd: Colloque du DIRO, vendredi le 2 février, 10:30, AA1360 Conférencier: William Hamilton
by Yoshua Bengio 01 Fév '18

01 Fév '18

Job talk for ML position. Candidate has a Stanford PhD. Please attend and give Simon L-J and I feedback. More candidates are coming over the next few weeks. ---------- Forwarded message ---------- From: "Pierre Poulin" <poulin(a)iro.umontreal.ca> Date: Jan 31, 2018 11:39 Subject: Colloque du DIRO, vendredi le 2 février, 10:30, AA1360 Conférencier: William Hamilton To: <seminaires(a)iro.umontreal.ca> Cc: Machine Learning for Computational Social Science par William Hamilton PhD candidate, Stanford University Vendredi 2 février, 10:30 à 11:30, salle *** 1360 ***, pavillon André-Aisenstadt Université de Montréal, 2920 chemin de la Tour Résumé: The combination of machine learning and massive social datasets has the potential to revolutionize our ability to predict and understand human behavior. However, machine learning on large social datasets is difficult because these datasets tend to be noisy, dynamic, and involve graph-structured relationships (e.g., social networks between users)—while traditional machine learning tools are largely designed for static datasets comprised of simple Euclidean vectors or grids. In this talk, I will describe new methods that I have developed for machine learning on massive, graph-structured social datasets. The technical focus of the talk will be on techniques for graph embedding, i.e., representation learning on graph-structured data. In the first part, I will describe how I have used graph embedding techniques to enable diverse social applications—from modeling cultural change to predicting conflict between online communities. In the second part, I will describe a new graph embedding framework, called GraphSAGE, that can scale to datasets that are orders of magnitude larger than previous approaches and that has been deployed at the website Pinterest to power a recommender system serving over 200 million users. I will close the talk with a general outlook on social AI technologies, including future technical directions and important ethical considerations. Bio: William (Will) Hamilton is a PhD Candidate in Computer Science at Stanford University, working jointly in the NLP and SNAP groups. He is co-advised by Dan Jurafsky and Jure Leskovec, and his interests lie at the intersection of machine learning, network science, natural language processing, and computational social science. Will's research is supported by the SAP Stanford Graduate Fellowship and an NSERC PGS-D Grant. Prior to coming to Stanford, Will completed a BSc and MSc at McGill University, where he studied machine learning under the supervision of Joelle Pineau and was the 2014 recipient of the Canadian AI Master's Thesis Award. Note: La présentation sera donnée en anglais. Pierre -- Pierre Poulin (poulin(a)iro.umontreal.ca) Directeur Dept. I.R.O., Universite de Montreal (normal mail) (courier) C.P. 6128, Succ. Centre-Ville 2920, chemin de la Tour, room 2163/2194 Montreal (Quebec) Canada, H3C 3J7 Montreal (Quebec) Canada, H3T 1J4 (514) 343-6780, (514) 343-5834 (fax) http://www.iro.umontreal.ca/~poulin

2 1

FW: [Tea Talk] Bart Van Parys (MIT) Fri Jan 26 1:30PM AA1360
by Daniel Doyon 29 Jan '18

29 Jan '18

Hi, I unfortunately missed Bart Van Parys’s Tea Talk last Friday, Jan26. I was curious if his presentation could be forwarded my way to get a flavor of the talk! Thanks, Dan From: lisa_teatalk-bounces(a)iro.umontreal.ca <mailto:lisa_teatalk-bounces@iro.umontreal.ca> [mailto:lisa_teatalk-bounces@iro.umontreal.ca] On Behalf Of Michael Noukhovitch Sent: Friday, January 26, 2018 1:11 PM To: Tea Talks MILA <teatalk-orgs(a)iro.umontreal.ca <mailto:teatalk-orgs@iro.umontreal.ca> >; lisa_teatalk(a)iro.umontreal.ca <mailto:lisa_teatalk@iro.umontreal.ca> ; lisa_seminaires(a)iro.umontreal.ca <mailto:lisa_seminaires@iro.umontreal.ca> ; Lisa Labo <lisa_labo(a)iro.umontreal.ca <mailto:lisa_labo@iro.umontreal.ca> >; lisa_montreal(a)iro.umontreal.ca <mailto:lisa_montreal@iro.umontreal.ca> Subject: Re: [Lisa_teatalk] [Tea Talk] Bart Van Parys (MIT) Fri Jan 26 1:30PM AA1360 Reminder: this is in 20 minutes! On Mon, Jan 22, 2018 at 3:35 PM Michael Noukhovitch <mnoukhov(a)gmail.com <mailto:mnoukhov@gmail.com> > wrote: Hi all! We will have a researcher from MIT, Bart Van Parys, giving a talk on Friday Jan 26th at 1:30PM in room AA1360. Note the later time, 1:30! This is a seminar is organized in collaboration with the Canada Research Chair in Decision Making under Uncertainty at GERAD! So this is ideal for all the optimization fans out there! Based on my experience, this will be a great talk and coming to it should be an easy decision! Michael KEYWORDS Optimization, Sparse Regression, Generalization TITLE Modern Optimization for Sparse Learning and Robust Analytics ABSTRACT We discuss the tremendous potential of integer optimization methods for learning predictive models from high-dimensional data via exact sparse regression. We show that novel integer formulations can solve exact sparse regression problems of sizes counting p=100,000s covariates for n=10,000s of samples. That is, two orders of magnitude more than current state of the art methods. We also indicate that robust optimization methods can help practitioners make data-driven decisions which are safeguarded against over-calibration to one particular data set. We claim that robust optimization methods have an enormous untapped potential when making subsequent decisions based on data. BIO Bart Van Parys is currently a postdoctoral researcher working with Prof. Dimitris Bertsimas at the MIT Sloan School of Management. His research interests are situated on the interface between optimization and machine learning. In 2015 he obtained his Ph.D. in control theory at the Swiss Federal Institute of Technology (ETH) in Zurich under the supervision of Prof. Manfred Morari. He received his M.E. from the University of Leuven in 2011.

1 0