{"id":10981,"date":"2018-12-27T13:27:28","date_gmt":"2018-12-27T18:27:28","guid":{"rendered":"https:\/\/vieux.ivado.ca?page_id=10981"},"modified":"2020-06-22T13:33:06","modified_gmt":"2020-06-22T17:33:06","slug":"excellence-scholarships-msc","status":"publish","type":"page","link":"https:\/\/vieux.ivado.ca\/en\/ivado-scholarships\/excellence-scholarships-msc\/","title":{"rendered":"Excellence Scholarships &#8211; Msc"},"content":{"rendered":"<h3 style=\"text-align: center;\"><img loading=\"lazy\" decoding=\"async\" class=\"aligncenter size-full wp-image-7984\" src=\"https:\/\/vieux.ivado.ca\/wp-content\/uploads\/2017\/10\/BanniereExcellenceScholarships.png\" alt=\"\" width=\"2189\" height=\"437\" srcset=\"https:\/\/vieux.ivado.ca\/wp-content\/uploads\/2017\/10\/BanniereExcellenceScholarships.png 2189w, https:\/\/vieux.ivado.ca\/wp-content\/uploads\/2017\/10\/BanniereExcellenceScholarships-300x60.png 300w, https:\/\/vieux.ivado.ca\/wp-content\/uploads\/2017\/10\/BanniereExcellenceScholarships-768x153.png 768w, https:\/\/vieux.ivado.ca\/wp-content\/uploads\/2017\/10\/BanniereExcellenceScholarships-1024x204.png 1024w\" sizes=\"(max-width: 2189px) 100vw, 2189px\" \/><\/h3>\n<p style=\"text-align: center;\">Next planned opening: Fall 2020<\/p>\n<h3 style=\"text-align: center;\">IVADO excellence scholarship<strong> program for Msc<\/strong><\/h3>\n<p><strong>IVADO\u2019s commitment to equity, diversity and inclusion and note to applicants<\/strong><br \/>\nTo ensure all members of society draw equal benefit from the advancement of knowledge and opportunities in data science, IVADO promotes equity, diversity and inclusion through each of its programs. IVADO aims to provide a recruitment process and research setting that are inclusive, non-discriminatory, open and transparent.<\/p>\n<div class=\"gdlr-shortcode-wrapper\"><div class=\"gdlr-item gdlr-tab-item horizontal\" ><div class=\"tab-title-wrapper\" ><h4 class=\"tab-title active\" ><span>Overview<\/span><\/h4><h4 class=\"tab-title\" ><span>Description<\/span><\/h4><h4 class=\"tab-title\" ><span>FAQ<\/span><\/h4><h4 class=\"tab-title\" ><span>Application<\/span><\/h4><h4 class=\"tab-title\" ><span>Results - 2018 Contest<\/span><\/h4><h4 class=\"tab-title\" ><span>Results - 2019 Contest<\/span><\/h4><h4 class=\"tab-title\" ><span>Results - 2020 Contest<\/span><\/h4><\/div><div class=\"tab-content-wrapper\" ><div class=\"tab-content active\" ><h3 class=\"gdlr-heading-shortcode \"  style=\"color: #000000;font-size: 16px;font-weight: bold;\" >Program description<\/h3>\n<ul>\n<li style=\"font-weight: 400;\"><span style=\"font-weight: 400;\"><strong>Field of study<\/strong>: IVADO Excellence scholarship program for Msc supports research on the <\/span><a href=\"http:\/\/www.cfref-apogee.gc.ca\/results-resultats\/abstracts-resumes\/competition_2\/universite_de_montreal-eng.aspx\" target=\"_blank\" rel=\"noopener noreferrer\"><span style=\"font-weight: 400;\">issues raised in the Canada First funding competition<\/span><\/a><span style=\"font-weight: 400;\">: data science in a broad sense, including methodological research in data science (machine learning, operations research, statistics) and its application in a range of sectors including the priority sectors of IVADO (health, transportation, logistics, energy, business and finance) or any other sector of application (sociology, physics, linguistics, engineering, etc.).<\/span><\/li>\n<li style=\"font-weight: 400;\"><span style=\"font-weight: 400;\"><strong>Amount of award and grant period<\/strong>:&nbsp;$20&nbsp;000 per year for a maximum of 6 sessions or 2 years<\/span><\/li>\n<li style=\"font-weight: 400;\"><b>Opening of the application process<\/b><span style=\"font-weight: 400;\">: January 9th, 2020&nbsp;9 a.m. EST<\/span><\/li>\n<li style=\"font-weight: 400;\"><b>Application deadline<\/b><span style=\"font-weight: 400;\">: February 17th, 2020&nbsp;9 a.m. EST<\/span><\/li>\n<li style=\"font-weight: 400;\"><b>Expected results notification date<\/b><span style=\"font-weight: 400;\"><span style=\"color: #000000;\">:<\/span> end of March 2020<\/span><\/li>\n<li style=\"font-weight: 400;\"><b>Criteria<\/b><span style=\"font-weight: 400;\">: See the description tab<\/span><\/li>\n<li style=\"font-weight: 400;\"><b>Submission<\/b><span style=\"font-weight: 400;\">: See the submission tab<\/span><\/li>\n<li><strong>Information<\/strong>: <a href=\"mailto:programmes-excellence@ivado.ca\">programmes-excellence@ivado.ca<\/a><\/li>\n<\/ul>\n<\/div><div class=\"tab-content\" ><p><span style=\"font-weight: 400;\"><\/p>\n<h3 class=\"gdlr-heading-shortcode \"  style=\"color: #000000;font-size: 16px;font-weight: bold;\" >Program objectives<\/span><\/h3>\n<p><span style=\"font-weight: 400;\">The goal of the excellence scholarship program is to support promising students in their training as future highly qualified personnel (researchers, professors, professionals) and more generally, future actors in the field of data science, mainly in IVADO members\u2019 areas of excellence: operations research, machine learning, decision sciences.<\/span><\/p>\n<h3 class=\"gdlr-heading-shortcode \"  style=\"color: #000000;font-size: 16px;font-weight: bold;\" >Eligibility<\/h3>\n<ul>\n<li style=\"font-weight: 400;\"><span style=\"font-weight: 400;\">Scholarship applicants must:<\/span>\n<ul>\n<li style=\"font-weight: 400;\"><span style=\"font-weight: 400;\">have already earned their Bachelor degree prior to their application date or intend to earn it by the date on which the competition results are announced. IVADO will be flexible with applicants who provide an adequate explanation for a career interruption or particular circumstances. This explanation must be included in the application (e.g. pregnancy\/maternity or sick leave);<\/span><\/li>\n<li style=\"font-weight: 400;\"><span style=\"font-weight: 400;\">intend to attend <strong>HEC Montr\u00e9al, Polytechnique Montr\u00e9al, Universit\u00e9 de Montr\u00e9al, McGill University or University of Alberta<\/strong>;<\/span><\/li>\n<li style=\"font-weight: 400;\"><span style=\"font-weight: 400;\">have a first class minimum average grade (3.7\/4.3 or 3.5\/4.00) over the previous years of study.<\/span><\/li>\n<\/ul>\n<\/li>\n<li style=\"font-weight: 400;\"><span style=\"font-weight: 400;\">Professor (supervisor) applicants must:<\/span>\n<ul>\n<li style=\"font-weight: 400;\"><span style=\"font-weight: 400;\">hold a faculty position as a professor at <strong>HEC Montr\u00e9al, Polytechnique Montr\u00e9al or Universit\u00e9 de Montr\u00e9al<\/strong>;<\/span><\/li>\n<li style=\"font-weight: 400;\"><span style=\"font-weight: 400;\">Professors at the <strong>University of Alberta<\/strong> and <strong>McGill University<\/strong> may act as supervisors providing they are full members of\u00a0 at least one of IVADO research group (Mila, CIRRELT, GERAD, CERC Data Science for real-time decision making, CRM, Tech3Lab).<\/span><\/li>\n<li>Eligible professors are assistant professors, associate professors, full professors, research professors or visiting professors. Adjunct professors are not eligible.<\/li>\n<li style=\"font-weight: 400;\"><b>Only submit one application to the competition.<\/b><\/li>\n<\/ul>\n<\/li>\n<li>For the co-supervisor, there is no constraint.<\/li>\n<\/ul>\n<h3 class=\"gdlr-heading-shortcode \"  style=\"color: #000000;font-size: 16px;font-weight: bold;\" >Funding period<\/h3>\n<p>The funding period starts in April 1st, 2020 or September 1st, 2020.<\/p>\n<h3 class=\"gdlr-heading-shortcode \"  style=\"color: #000000;font-size: 16px;font-weight: bold;\" >Amounts and terms <\/h3>\n<p><span style=\"font-weight: 400;\">The funds shall be transferred to the office of research of the supervisor\u2019s university, and the university shall pay the student according to its own compensation rules. For projects that require ethics approval, the funds shall only be paid out once the approval is granted. Some projects may require specific agreements (e.g. pertaining to intellectual property).<\/span><\/p>\n<p><span style=\"font-weight: 400;\">Funding may be cut, withheld, delayed or rescinded under the circumstances outlined in the letter of award.<\/span><\/p>\n<h3 class=\"gdlr-heading-shortcode \"  style=\"color: #000000;font-size: 16px;font-weight: bold;\" >Competitive process <\/h3>\n<p><strong>Review and criteria<\/strong><\/p>\n<p><span style=\"font-weight: 400;\">The applications shall be reviewed to ensure compliance with program rules (e.g. applications that are incomplete, exceed the page limit or list an ineligible applicant or supervisor). Only applications that meet all criteria will be forwarded to the review committee.<\/span><\/p>\n<p><span style=\"font-weight: 400;\">The parity-based review committee shall be made up of university professors and shall not be listed as a supervisor by any applicant. However, given the small size of the communities in certain areas, it may prove difficult to select expert reviewers who are not included in an application submitted to the competition. In such cases, a reviewer may be required to assess an application despite being listed in another application as a supervisor. An external reviewer may also join the committee. The committee shall ensure by all possible means that the reviewer does not influence the ranking of the application in which he\/she is included.<\/span><\/p>\n<p><span style=\"font-weight: 400;\">The review committee will check the project\u2019s alignment between the research project and IVADO\u2019s scientific direction, then shall rank the applications based on <\/span><b>excellence<\/b><span style=\"font-weight: 400;\">, as well as the project\u2019s <\/span><b>alignment with IVADO\u2019s overarching framework<\/b><span style=\"font-weight: 400;\">, which aims to promote multidisciplinary collaboration and diversity in data science.\u00a0<\/span><\/p>\n<p><span style=\"font-weight: 400;\">In terms of excellence, the committee will specifically assess:<\/span><\/p>\n<ul>\n<li style=\"font-weight: 400;\"><span style=\"font-weight: 400;\">Research ability<\/span><\/li>\n<li style=\"font-weight: 400;\"><span style=\"font-weight: 400;\">Depth and breadth of experience: multidisciplinary and professional experiences, extra-academic activities, collaborations, contributions to the scientific community and society as a whole, etc.<\/span><\/li>\n<li style=\"font-weight: 400;\"><span style=\"font-weight: 400;\">Expected adequacy with the proposed project<\/span><\/li>\n<\/ul>\n<p><strong>Final step and commitments<\/strong><\/p>\n<p><span style=\"font-weight: 400;\">The student shall:<\/span><\/p>\n<ul>\n<li style=\"font-weight: 400;\"><span style=\"font-weight: 400;\">be physically present at his\/her supervisor\u2019s university;<\/span><\/li>\n<li style=\"font-weight: 400;\"><span style=\"font-weight: 400;\">contribute to IVADO\u2019s community and activities by, for example, taking part in:<\/span>\n<ul>\n<li style=\"font-weight: 400;\"><span style=\"font-weight: 400;\">presentations on his\/her research;<\/span><\/li>\n<li style=\"font-weight: 400;\"><span style=\"font-weight: 400;\">training and knowledge dissemination activities;<\/span><\/li>\n<li style=\"font-weight: 400;\"><span style=\"font-weight: 400;\">consultations;<\/span><\/li>\n<li style=\"font-weight: 400;\"><span style=\"font-weight: 400;\">activities generally undertaken by career researchers (mentorship, assessment, co-organization of events, etc.);<\/span><\/li>\n<\/ul>\n<\/li>\n<li style=\"font-weight: 400;\"><span style=\"font-weight: 400;\">recognize that he\/she is a member of an academic community to which he\/she shall contribute;<\/span><\/li>\n<li style=\"font-weight: 400;\"><span style=\"font-weight: 400;\">comply with the <\/span><a href=\"http:\/\/www.science.gc.ca\/eic\/site\/063.nsf\/fra\/h_F6765465.html\" target=\"_blank\" rel=\"noopener noreferrer\"><span style=\"font-weight: 400;\">Tri-Agency Open Access Policy on Publication<\/span><\/a><span style=\"font-weight: 400;\">. Students are encouraged to publish their research findings (papers, recordings of presentations, source codes, databases, etc.), in compliance with the intellectual property rules that apply to their own specific case;<\/span><\/li>\n<li style=\"font-weight: 400;\"><span style=\"font-weight: 400;\">recognize the financial support granted by IVADO and the CFREF or FRQ when disseminating the research results and, more broadly, in all the activities in which he\/she takes part<\/span><\/li>\n<\/ul>\n<p><span style=\"font-weight: 400;\">The supervisor shall:<\/span><\/p>\n<ul>\n<li style=\"font-weight: 400;\"><span style=\"font-weight: 400;\">provide a work environment that is conducive to the completion of the project<\/span><\/li>\n<li style=\"font-weight: 400;\"><span style=\"font-weight: 400;\">oversee the work of the student<\/span><\/li>\n<\/ul>\n<\/div><div class=\"tab-content\" ><h3 class=\"gdlr-heading-shortcode \"  style=\"color: #00000;font-size: 16px;font-weight: bold;\" >FAQ<\/h3>\n<ul>\n<li style=\"font-weight: 400;\"><span style=\"font-weight: 400;\">Is there a particular format for preparing a CV?<\/span>\n<ul>\n<li style=\"font-weight: 400;\"><span style=\"font-weight: 400;\">No, there is no particular format that needs to be followed. However, each piece of the record must help the assessor to form an opinion on the record. A CV that is too long or confusing may make evaluation more difficult.<\/span><\/li>\n<\/ul>\n<\/li>\n<li style=\"font-weight: 400;\"><span style=\"font-weight: 400;\">Are there any specific rules for the recommendation letter?<\/span>\n<ul>\n<li style=\"font-weight: 400;\"><span style=\"font-weight: 400;\">No, there are no specific rules for the recommendation letter.<\/span><\/li>\n<\/ul>\n<\/li>\n<li style=\"font-weight: 400;\"><span style=\"font-weight: 400;\">Can candidates send recommendation letters themselves?<\/span>\n<ul>\n<li style=\"font-weight: 400;\"><span style=\"font-weight: 400;\"> No, recommendation letters can only be upload by their author in the platform.<\/span><\/li>\n<\/ul>\n<\/li>\n<li style=\"font-weight: 400;\"><span style=\"font-weight: 400;\">Can I send my unofficial transcript?<\/span>\n<ul>\n<li style=\"font-weight: 400;\"><span style=\"font-weight: 400;\">No, you must upload your official transcript including all your current results. Originals or certified copies must be scanned and uploaded to the application and for non-Canadian universities, you must specify the rating scale.<\/span><\/li>\n<\/ul>\n<\/li>\n<\/ul>\n<p>Didn\u2019t find what you were looking for? If not, send us an <a href=\"mailto:programmes-excellence@ivado.ca\">e-mail<\/a>.<\/p>\n<\/div><div class=\"tab-content\" ><p><strong>Please apply through:\u00a0<a href=\"https:\/\/ivado.smapply.io\/\">https:\/\/ivado.smapply.io\/<\/a><\/strong><\/p>\n<p>All applications sent by email will be rejected.<\/p>\n<p>All applications must contain:<\/p>\n<ul>\n<li>a questionnaire to be completed on the platform WITH a common-language description of the project (maximum length of one page);\n<ul>\n<li>an example of filled form\u00a0is <a href=\"https:\/\/vieux.ivado.ca\/wp-content\/uploads\/2020\/01\/Maitrise2020_Formulaire-test.pdf\" target=\"_blank\" rel=\"noopener noreferrer\">available here<\/a> <span style=\"font-weight: 400;\">and <\/span><a href=\"https:\/\/vieux.ivado.ca\/wp-content\/uploads\/2019\/03\/SelfIdentificationPhD_2019.pdf\" target=\"_blank\" rel=\"noopener noreferrer\">self-declaration form.<\/a><\/li>\n<\/ul>\n<\/li>\n<li>student CV (free format) to be uploaded;<\/li>\n<li>Bachelor official transcripts and Master\u2019s marks if already started (as well as information on the grading scale when the transcript is issued by a non-Canadian university);<\/li>\n<li>Two recommendation letters, including a letter directly uploaded by your Msc supervisor (actual or potential future).<\/li>\n<\/ul>\n<\/div><div class=\"tab-content\" ><ul>\n<li><strong>Larry Dong<\/strong> (McGill University, Erica Moodie)\n<ul>\n<li>When making decisions, medical professionals often rely on past experience and their own judgment. However, it is often the case that an individual decision-makerfaces a situation that is unfamiliar to him or her. An adaptive treatment strategy (ATS) can help such biomedical experts in their decision-making, as they are a statistical representation of a decision algorithm for a given treatment that optimizes patient outcomes. ATSs are estimated with large amounts of data, but an issue that may occur is that such sources of data may be subject to unmeasured confounding, whereby important variables needed to ensure the causal inference are missing. The idea behind this research project is to develop a sensitivity analysis to better understand and to quantify the impact of unmeasured confounding on decision rules in ATSs.<\/li>\n<\/ul>\n<\/li>\n<li><strong>Jonathan Pilault<\/strong> (Polytechnique Montr\u00e9al, Christopher Pal)\n<ul>\n<li>Language understanding and generation is a unique capacity of humans. Automatic summarization is an important task in Natural (human) Language Processing. This task consists in reducing the size of discourse while preserving information content. Abstractive summarization sets itself apart from other types of summarization since it most closely relates to how humans would summarize a book, a movie, an article or a conversation. From a research standpoint, automatic abstractive summarization is interesting since it requires models to both understand and generate human language. In the past year, we have seen research that have improved the ability of Neural Networks to choose the most important parts of discourse while beginning to address key pain points (e.g. repeating sentences, nonsensical formulations) during summary text generation. Recent techniques in Computer Vision image generation tasks have shown that image quality can be further improved using Generative Adversarial Networks (GAN). Our intuition is that the same is true for a Natural Language Processing task. We propose to incorporate newest GAN architectures into some of the most novel abstractive summarization models to validate our hypothesis. The objective is to create a state-of-the-art summarization system that most closely mimics human summarizers. This outcome will also bring us closer to understand GANs analytically.<\/li>\n<\/ul>\n<\/li>\n<li><strong>Alice Wu<\/strong> (Polytechnique Montr\u00e9al, Fran\u00e7ois Soumis)\n<ul>\n<li>Combiner l\u2019A.I. et la R.O. pour optimiser les blocs mensuels d\u2019\u00e9quipages a\u00e9rien.Nos travaux r\u00e9cents portent sur le d\u00e9veloppement de deux nouveaux algorithmes Improved Primal Simplex (IPS) et Integral Simplex Using Decomposition (ISUD) qui profitent de l\u2019information a priori sur les solutions attendues pour r\u00e9duire le nombre de variables et de contraintes \u00e0 traiter simultan\u00e9ment. Actuellement cette information est donn\u00e9e par des r\u00e8gles fournies par les planificateurs. L\u2019objectif de recherche sera de d\u00e9velopper un syst\u00e8me utilisant l\u2019intelligence artificielle (IA) pour estimer la probabilit\u00e9 que la variable liant deux rotations fasse partie de la solution d\u2019un probl\u00e8me de blocs mensuels d\u2019\u00e9quipages a\u00e9riens. L\u2019apprentissage se fera sur les donn\u00e9es historiques de plusieurs mois, de plusieurs types d\u2019avions et de plusieurs compagnies. L\u2019estimation des probabilit\u00e9s doit se faire \u00e0 partir des caract\u00e9ristiques des rotations et non \u00e0 partir de leurs noms. Une rotation ne revient pas d\u2019une compagnie \u00e0 l\u2019autre ni d\u2019un mois \u00e0 l\u2019autre. Il faudra identifier les caract\u00e9ristiques pertinentes. Il faudra de la recherche sur l\u2019apprentissage pour profiter des contraintes du probl\u00e8me. Il y a des contraintes entre le personnel terminant des rotations et celui en commen\u00e7ant par la suite. La validation de l\u2019apprentissage se fera en alimentant les optimiseurs avec l\u2019information estim\u00e9e et en observant la qualit\u00e9 des solutions obtenues et les temps de calcul. Il y aura de la recherche \u00e0 faire dans les optimiseurs pour exploiter au mieux cette nouvelle information.<\/li>\n<\/ul>\n<\/li>\n<\/ul>\n<\/div><div class=\"tab-content\" ><p><strong>Tiphaine Bonniot de Ruisselet<\/strong> (Polytechnique Montr\u00e9al, Dominique Orban)<\/p>\n<ul>\n<li>Acc\u00e9l\u00e9ration de m\u00e9thodes d\u2019optimisation pour les probl\u00e8mes volumineux par \u00e9valuation inexact<\/li>\n<\/ul>\n<p>Nous nous int\u00e9ressons aux probl\u00e8mes d&#8217;optimisation continue, non convexe et sans contraintes dans lesquels l&#8217;\u00e9valuation des valeurs de l&#8217;objectif et de son gradient sont obtenues \u00e0 l&#8217;issue d&#8217;un processus co\u00fbteux. Nous supposons que l&#8217;on peut obtenir \u00e0 moindre co\u00fbts des approximations de l&#8217;objectif et de son gradient \u00e0 un niveau de pr\u00e9cision souhait\u00e9. Nous regarderons l&#8217;impact de ces hypoth\u00e8ses sur la convergence et la complexit\u00e9 de m\u00e9thodes d&#8217;optimisation classiques ainsi que les \u00e9conomies pouvant \u00eatre effectu\u00e9es sur le temps de calcul et la consommation \u00e9nerg\u00e9tique. Cette \u00e9tude est motiv\u00e9e, entre autre, par les probl\u00e8mes d&#8217;inversion sismique dont la taille peut avoisiner les centaines de millions de variables et dont la fonction et le gradient peuvent \u00eatre approxim\u00e9s par la r\u00e9solution d&#8217;un probl\u00e8me aux moindres carr\u00e9s lin\u00e9aires. L&#8217;\u00e9conomie de temps de calcul et d&#8217;\u00e9nergie est un enjeu majeur de l&#8217;\u00e8re de l&#8217;intelligence artificielle et de l&#8217;exploration des donn\u00e9es volumineuses et cette approche est nouvelle est prometteuse en terme de retomb\u00e9es \u00e9conomiques et environnementales.<\/p>\n<p><strong>Stephanie Cairns<\/strong> (McGill University, Adam Oberman)<\/p>\n<ul>\n<li>Oberman Mathematical approaches to adversarial robustness and confidence in DNN<\/li>\n<\/ul>\n<p>Deep convolutional neural networks are highly effective at image classification tasks, achieving higher accuracy than conventional machine learning methods but lacking the performance guarantees associated with these methods. Without additional performance guarantees, for example error bounds, they cannot be safely used in applications where errors can be costly. There is a consensus amongst researchers that greater interpretability and robustness are needed. Robustness can be to differences in the data set where the models can be deployed, or even robustness to adversarial samples: perturbations of the data designed deliberately by an adversary to lead to a misclassification.<\/p>\n<p>In this project, we will study reliability in two contexts: (i) developing improved confidence in the prediction of the neural network, using modified losses to improve confidence measures (ii) modified losses which result in better robustness to adversarial examples. The overall goal of the project is to lead to more reliable deep learning models.<\/p>\n<p><strong>Enora Georgeault<\/strong> (HEC Montr\u00e9al, Marie-\u00c8ve Rancourt)<\/p>\n<ul>\n<li>Mod\u00e8les pr\u00e9dictifs de l\u2019allocation des dons de la Croix-Rouge canadienne en r\u00e9ponse aux feux de for\u00eat<\/li>\n<\/ul>\n<p>Au Canada, les inondations et les feux de for\u00eat sont les catastrophes naturelles qui provoquent le plus de d\u00e9g\u00e2ts. Les efforts de la Croix-Rouge canadienne (CRC) visant \u00e0 att\u00e9nuer les impacts des feux de for\u00eat d\u00e9pendent fortement de la capacit\u00e9 des organisations \u00e0 planifier, \u00e0 l\u2019avance, les op\u00e9rations logistiques de secours. Le premier objectif du projet est l\u2019\u00e9laboration de mod\u00e8les permettant de pr\u00e9dire l\u2019allocation des dons en argent aux b\u00e9n\u00e9ficiaires, selon les caract\u00e9ristiques socio-d\u00e9mographiques de la r\u00e9gion et du b\u00e9n\u00e9ficiaire ainsi que selon les caract\u00e9ristiques des feux (s\u00e9v\u00e9rit\u00e9 et type). Le second objectif sera de comprendre les facteurs qui ont un impact significatif sur les besoins de la CRC lors d\u2019une r\u00e9ponse \u00e0 un feu de for\u00eat, afin de faciliter la planification des op\u00e9rations logistiques et les appels de financement.<\/p>\n<p><strong>Bhargav Kanuparthi<\/strong> (Universit\u00e9 de Montr\u00e9al, Yoshua Bengio)<\/p>\n<ul>\n<li>h detach Modifying the LSTM Gradient Towards Better Optimization<\/li>\n<\/ul>\n<p>Recurrent neural networks are known for their notorious exploding and vanishing gradient problem (EVGP). This problem becomes more evident in tasks where the information needed to correctly solve them exist over long time scales, because EVGP prevents important gradient components from being back-propagated adequately over a large number of steps. We introduce a simple stochastic algorithm (\\textit{h}-detach) that is specific to LSTM optimization and targeted towards addressing this problem. Specifically, we show that when the LSTM weights are large, the gradient components through the linear path (cell state) in the LSTM computational graph get suppressed. Based on the hypothesis that these components carry information about long term dependencies (which we show empirically), their suppression can prevent LSTMs from capturing them. Our algorithm\\footnote{Our code is available at https:\/\/github.com\/bhargav104\/h-detach.} prevents gradients flowing through this path from getting suppressed, thus allowing the LSTM to capture such dependencies better. We show significant improvements over vanilla LSTM gradient based training in terms of convergence speed, robustness to seed and learning rate, and generalization using our modification of LSTM gradient on various benchmark datasets.<\/p>\n<p><strong>Vincent Labont\u00e9<\/strong> (Polytechnique Montr\u00e9al, Michel Gagnon)<\/p>\n<ul>\n<li>Extraction de connaissances en fran\u00e7ais bas\u00e9e sur une traduction des textes en anglais combin\u00e9e \u00e0 l\u2019utilisation d\u2019outils d\u00e9velopp\u00e9s pour l\u2019anglais<\/li>\n<\/ul>\n<p>Plusieurs institutions gouvernementales rendent disponible sur leurs sites web un tr\u00e8s grand volume de documents qui ne sont \u00e9crits que dans la langue officielle du pays. Or, de plus en plus ces institutions d\u00e9sirent transformer ces documents en une base de connaissances, d\u00e9ploy\u00e9e en un ensemble de donn\u00e9es ouvertes int\u00e9gr\u00e9es au Web s\u00e9mantique. C\u2019est le cas notamment du minist\u00e8re de la Culture et des Communications du Qu\u00e9bec, qui met \u00e0 la disposition du public un r\u00e9pertoire du patrimoine culturel du Qu\u00e9bec, tr\u00e8s riche en informations textuelles, mais qu\u2019il est malheureusement difficile d\u2019int\u00e9grer aux donn\u00e9es des autres acteurs culturels du Qu\u00e9bec, ou de lier \u00e0 toutes les connaissances patrimoniales qui sont d\u00e9j\u00e0 pr\u00e9sentes dans le r\u00e9seau de donn\u00e9es ouvertes Linked Open Data (LOD).<\/p>\n<p>Plusieurs travaux ont d\u00e9j\u00e0 \u00e9t\u00e9 propos\u00e9s pour soutenir l\u2019effort d\u2019extraction de connaissances \u00e0 partir de textes&nbsp;: des annotateurs s\u00e9mantiques, qui identifient dans un document les entit\u00e9s qui y sont cit\u00e9es (personnes, organisations, etc.) et les lient \u00e0 leur repr\u00e9sentation dans une base de connaissances du LOD; des extracteurs de relations, capables d\u2019extraire du texte des relations entre deux entit\u00e9s (par exemple, \u00ab&nbsp;X est l\u2019auteur du roman Y&nbsp;\u00bb); des extracteurs d\u2019\u00e9v\u00e9nements et d\u2019informations temporelles. Dans la tr\u00e8s grande majorit\u00e9 des cas, ces outils ont \u00e9t\u00e9 d\u00e9velopp\u00e9s pour l\u2019anglais, ou offrent de pi\u00e8tres performances lorsqu\u2019appliqu\u00e9s au fran\u00e7ais.<\/p>\n<p>Nous proposons donc d\u2019explorer une approche qui consiste \u00e0 produire, \u00e0 partir d\u2019un corpus de documents en fran\u00e7ais, une version \u00e9quivalente traduite sur laquelle seront appliqu\u00e9s les outils d\u00e9j\u00e0 existants pour l\u2019anglais (le service Syntaxnet de Google, par exemple). Cela implique qu\u2019il faudra tenir compte des erreurs et inexactitudes qui r\u00e9sulteront de l\u2019\u00e9tape de traduction. Pour y arriver, des techniques de paraphrase et de simplification de texte seront explor\u00e9es, l\u2019hypoth\u00e8se ici \u00e9tant que des phrases simples sont plus faciles \u00e0 traduire et que cette simplification n\u2019aura pas d\u2019impact majeur sur la r\u00e9solution de la t\u00e2che si la s\u00e9mantique est pr\u00e9serv\u00e9e lors de cette simplification. On notera aussi que certains aspects de la langue, comme l\u2019anaphore, perturbent la traduction (le module de traduction aura du mal \u00e0 choisir entre les pronoms \u00ab&nbsp;it&nbsp;\u00bb et \u00ab&nbsp;he&nbsp;\u00bb pour traduire le pronom \u00ab&nbsp;il&nbsp;\u00bb). Il faudra dans ces cas mesurer pr\u00e9cis\u00e9ment leur impact et proposer des solutions de contournement.<\/p>\n<p>En bref, le projet propos\u00e9 permettra de d\u00e9terminer dans quelle mesure les services de traduction actuellement disponibles pr\u00e9servent suffisamment le sens du texte pour pouvoir exploiter des outils d\u00e9velopp\u00e9s pour une autre langue. L\u2019hypoth\u00e8se que nous d\u00e9sirons valider est que leurs lacunes peuvent \u00eatre combl\u00e9es par certains pr\u00e9traitements du texte original, et que ces pr\u00e9traitements peuvent \u00eatre impl\u00e9ment\u00e9e \u00e0 faibles co\u00fbts (en temps et en ressources).<\/p>\n<p><strong>Thomas MacDougall<\/strong> (Universit\u00e9 de Montr\u00e9al, S\u00e9bastien Lemieux)<\/p>\n<ul>\n<li>Use of Deep Learning Approaches in the Activity Prediction and Design of Therapeutic Molecules<\/li>\n<\/ul>\n<p>The proposed research is to employ Deep Learning and Neural Networks, which are both fields of Machine Learning, to more accurately predict the effectiveness, or \u201cactivity\u201d, of potential therapeutic molecules (potential drugs). We are primarily concerned with predicting a given molecule&#8217;s ability to inhibit the growth of primary patient cancer cells (cells taken directly from a patient). The Leucegene project at the Institut de Recherche en Immunologie et Canc\u00e9rologie (IRIC) has tested the activity of a large number of compounds in inhibiting the growth of cancer cells from patients afflicted with acute myeloid leukemia. The proposed research will use this activity data, along with several other data sources, to build an algorithm that can better predict the effectiveness that a molecule will have in inhibiting cancer cell growth. This means that before a molecule is even synthesized in a chemistry lab, a good estimation of its effectiveness as a therapeutic compound can be made, almost instantly. The first approach is to use Neural Networks and \u201crepresentation learning\u201d, in which features of the molecules that are important to improving activity are identified automatically by the algorithm. This will be done by representing the molecules as graphs and networks. Another approach that will be taken is the use of \u201cmulti-task learning\u201d in which the prediction accuracy of an algorithm can be improved if the same algorithm is trained for multiple tasks on multiple datasets. The &#8220;multiple tasks&#8221; that will be focused on are multiple, but related, drug targets that are essential to cancer cell growth. Moving beyond activity prediction alone, these machine learning architectures will be expanded to design new chemical structures for potential drug molecules, based on information that is learned from drug molecules with known activities. These approaches have the capacity to improve the predictions about whether molecules will make effective drugs, and to design new molecules that have even better effectiveness than known drugs. Research progress in this area will lower the cost, both in money and time, of the drug development process.<\/p>\n<p><strong>Bhairav Mehta<\/strong> (Universit\u00e9 de Montr\u00e9al, Liam Paull)<\/p>\n<ul>\n<li>Attacking the Reality Gap in Robotic Reinforcement Learning<\/li>\n<\/ul>\n<p>As Reinforcement Learning (RL) becomes an increasingly popular avenue of research, one area that stands to be revolutionized is robotics. However, one prominent downside of applying RL in robotics scenarios is the amount of experience today\u2019s RL algorithms require to learn. Since these data-intensive policies cannot be learned on real robots due to time constraints, researchers turn to fast, approximate simulators. Trading off accuracy for speed can cause problems at test time, and policies that fail to transfer to the real world fall prey to the reality gap: the differences between training simulation and the real-world robot. Our project focuses on theoretically analyzing this issue, and provides practical algorithms to improve safety and robustness when transferring robotic policies out of simulation. We propose algorithms that use expert-collected robot data to learn a simulator, allowing for better modeling of the testing distribution and minimizing the reality gap upon transfer. In addition, we study the transfer problem using analysis tools from dynamical systems and continual learning research, looking for indicators in neural network dynamics and optimization that signal when the reality gap is likely to pose an issue. Lastly, we use the analysis to synthesize an algorithm which optimizes for the metrics that signal good, \u201ctransferable\u201d policies, allowing safer and more robust sim-to-real transfer.<\/p>\n<p><strong>Timothy Nest<\/strong> (Universit\u00e9 de Montr\u00e9al, Karim Jerbi)<\/p>\n<ul>\n<li>Leveraging Machine Learning and Magnetoencephalography for the Study of Normal and atypical states of Consciousness<\/li>\n<\/ul>\n<p>Understanding the neural processes and network dynamics underlying conscious perception is a complex yet important challenge that lies at the intersection between cognitive brain imaging, mental health, and data science. Magnetoencephalography (MEG) is a brain imaging technique that has many qualities favorable to investigating conscious perception due to its high temporal resolution and high signal to noise ratio. However MEG analyses across space, time and frequency is challenging due to the extreme high-dimensionality of variables of interest, and susceptibility to overfitting. Furthermore, high-computational complexity limits the ease with which investigators might approach some cross-frequency coupling metrics believed to be important for conscious perception and integration, across the whole brain. To mitigate such challenges, researchers frequently rely on a variety of multivariate feature extraction and compression algorithms. However, these techniques still require substantial tuning, and are limited in their application to the kinds of high-order tensor structures encountered in MEG. New methods for the study of conscious perception with MEG are thus needed.<\/p>\n<p>In this project, we will leverage very recent advances in computer science and machine learning that extend algorithms currently used in neuroimaging research, to extreme high-dimensional spaces. Taken together, the proposed research will apply state-of-the-art techniques in machine-learning and electrophysiological signal processing to overcome current obstacles in the study of the brain processes that mediate conscious perception. This work will constitute an important contribution to neuroimaging methodology, neuropharmacology, and psychiatry. Beyond expanding our understanding of healthy cognition, this research may ultimately provide novel paths to the study of psychiatric disorders that involve altered conscious perception, such as Schizophrenia.<\/p>\n<p><strong>Jacinthe Pilette<\/strong> (Universit\u00e9 de Montr\u00e9al, Jean-Fran\u00e7ois Arguin)<\/p>\n<ul>\n<li>Recherche de nouvelle physique au Grand collisionneur de hadrons (LHC) \u00e0 l&#8217;aide de l&#8217;apprentissage profond<\/li>\n<\/ul>\n<p>Le Grand collisionneur de hadrons (LHC) se situe au c\u0153ur de la recherche fondamentale en physique. Avec sa circonf\u00e9rence de 27 km, celui-ci constitue le plus grand et plus puissant acc\u00e9l\u00e9rateur de particules au monde. Ceci en fait le meilleur outil afin d\u2019\u00e9tudier le domaine de l\u2019infiniment petit. C\u2019est d\u2019ailleurs au LHC que le boson de Higgs fut d\u00e9couvert, menant \u00e0 l\u2019obtention du prix Nobel de physique en 2013.<\/p>\n<p>Cependant, le mod\u00e8le standard, r\u00e9f\u00e9rence qui dicte les lois r\u00e9gissant les particules et leurs interactions, poss\u00e8de plusieurs lacunes que les physiciens et physiciennes n\u2019ont toujours pas r\u00e9ussi \u00e0 combler. Plusieurs th\u00e9ories furent \u00e9labor\u00e9es, mais aucune d\u2019entre elles ne fut observ\u00e9e au LHC. Face \u00e0 ce d\u00e9fi, la communaut\u00e9 de physique des particules devra utiliser une nouvelle approche.<\/p>\n<p>Le groupe ATLAS de l\u2019Universit\u00e9 de Montr\u00e9al s\u2019est ainsi tourn\u00e9 vers l\u2019intelligence artificielle. Le projet \u00e9labor\u00e9 par cette collaboration, et l\u2019objectif principal de cette recherche est de d\u00e9velopper un algorithme d\u2019apprentissage profond qui permettrait de d\u00e9tecter des anomalies dans les donn\u00e9es. L\u2019algorithme d\u00e9velopp\u00e9 sera ensuite utilis\u00e9 sur les donn\u00e9es du d\u00e9tecteur ATLAS dans l\u2019espoir de d\u00e9couvrir des signaux de nouvelle physique et d\u2019am\u00e9liorer notre compr\u00e9hension de l\u2019univers.<\/p>\n<p><strong>L\u00e9a Ricard<\/strong> (Universit\u00e9 de Montr\u00e9al, Emma Frejinger)<\/p>\n<ul>\n<li>Mod\u00e9lisation de la probabilit\u00e9 d\u2019acceptation d\u2019une route dans un contexte de covoiturage<\/li>\n<\/ul>\n<p>Le covoiturage touche aux algorithmes fr\u00e9quemment \u00e9tudi\u00e9s de tourn\u00e9es de v\u00e9hicule, de ramassage et de livraison avec fen\u00eatres de temps et de transport \u00e0 la demande dynamique. Toutefois, tr\u00e8s peu d\u2019\u00e9tudes s\u2019attardent au contexte o\u00f9 les conducteurs et les passagers peuvent rejeter une proposition de route. Alors que le rejet d\u2019une route propos\u00e9e est rare lorsque les conducteurs sont des professionnels, c\u2019est plut\u00f4t la norme dans un contexte de covoiturage. La mod\u00e9lisation de la probabilit\u00e9 d\u2019acceptation d\u2019une route se pose alors comme un probl\u00e8me central dans le d\u00e9veloppement d\u2019une application mobile de covoiturage de qualit\u00e9.<\/p>\n<p>Le mod\u00e8le d\u2019apprentissage automatique d\u00e9velopp\u00e9 devra estimer, selon les caract\u00e9ristiques de l\u2019utilisateur (notamment s\u2019il est conducteur ou passager) et les routes alternatives propos\u00e9es, la probabilit\u00e9 d\u2019acceptation d\u2019une route. De prime abord, cette mod\u00e9lisation pose deux d\u00e9fis&nbsp;:<\/p>\n<p>(1) La fa\u00e7on dont les acceptations et les refus sont collect\u00e9s pose un probl\u00e8me de type logged bandit. \u00c0 ce titre, plusieurs propositions peuvent \u00eatre offertes en m\u00eame temps et un utilisateur peut en accepter plusieurs. De plus, les offres peuvent \u00eatre activement refus\u00e9es, simplement ignor\u00e9es ou accept\u00e9es. Puisque les offres sont affich\u00e9es s\u00e9quentiellement, celles qui apparaissent en premier ont plus de chance d\u2019attirer l\u2019attention de l\u2019utilisateur. L\u2019ordre des propositions a donc probablement une influence sur la probabilit\u00e9 d\u2019acceptation.<br \/>\n(2) Le comportement des nouveaux utilisateurs, pour qui tr\u00e8s peu d\u2019information est disponible, devra \u00eatre inf\u00e9r\u00e9 \u00e0 partir des clients similaires de longue date. Ceci est en soi un probl\u00e8me difficile.<\/p>\n<p><strong>Alexandre Riviello<\/strong> (Polytechnique Montr\u00e9al, Jean-Pierre David)<\/p>\n<ul>\n<li>Hardware Acceleration of Speech Recognition Algorithms<\/li>\n<\/ul>\n<p>Speech recognition has become prevalent in our lives in recent years. Personal assistants, such as Amazon\u2019s Alexa or Apple\u2019s Siri are such examples. With the rise of deep learning, speech recognition algorithms gained a lot of precision. This is due, mostly, to the use of neural networks. These complex algorithms, used in the context of a classification task, can distinguish between different characters, phonemes or words. However, they require lots of computations, limiting their use in power-constrained devices, such as smartphones. In my research, I will attempt to find hardware-friendly implementations of these networks. Deep learning algorithms are usually written in high-level languages using frameworks such as Torch or Tensorflow. To generate hardware-friendly representations, models will be adapted, using these frameworks. For example, recent findings have shown that basic networks can use weights and activations represented over 1 or 2 bits and retain their accuracy. The reduction of the precision of the network parameters is called quantization. This concept will be one of the many ways used to simplify the networks. Another aspect of this research will be to revisit methods of representing voice features. Traditionally, spoken utterances were converted to Mel Frequency Cepstrum Coefficients (MFCCs) which are essentially values representing signal power over a frequency axis. These coefficients are calculated roughly every 10 ms and are then sent to the model network. A representation of lower precision can greatly reduce the computational costs of the network. The overall goal of the research will be to improve the calculation speed and to diminish the power consumption of speech recognition algorithms.<\/p>\n<\/div><div class=\"tab-content\" ><p><strong>Alexandre Adam<\/strong> (Universit\u00e9 de Montr\u00e9al, Laurence Perreault Levasseur)<\/p>\n<ul>\n<li>Mesurer l&#8217;expansion de l&#8217;Univers avec l&#8217;apprentissage automatique<\/li>\n<\/ul>\n<p style=\"padding-left: 40px;\">Le taux d\u2019expansion de l\u2019Univers est une observable importante pour contraindre les mod\u00e8les cosmologiques qui retracent l\u2019\u00e9volution de l\u2019Univers depuis le Big Bang. R\u00e9cemment (2018), l\u2019\u00e9quipe du satellite Planck a publi\u00e9 une valeur d\u00e9riv\u00e9e des mesures du rayonnement fossile \u00e9mis lorsque l&#8217;Univers n&#8217;\u00e9tait \u00e2g\u00e9 que de 300,000 ans. La valeur trouv\u00e9e contredit les mesures locales du param\u00e8tre, faites \u00e0 partir de la vitesse de fuite des supernovas Ia et des c\u00e9ph\u00e9ides se trouvant pr\u00e8s de la Voie lact\u00e9e. Nous proposons d&#8217;investiguer ce probl\u00e8me via une troisi\u00e8me m\u00e9thode de mesure qui, jusqu&#8217;\u00e0 maintenant, poss\u00e9dait une pr\u00e9cision limit\u00e9e par la faible quantit\u00e9 connue de quasar situ\u00e9 derri\u00e8re une galaxie selon notre ligne de vue, telle que l&#8217;image du quasar est multipli\u00e9e par l&#8217;effet de lentille gravitationnelle. La pr\u00e9cision de cette m\u00e9thode est limit\u00e9e en grande partie par la reconstruction de la distribution de masse de la galaxie-lentille. Les avanc\u00e9es r\u00e9centes des algorithmes d\u2019apprentissage automatiques ont permis de d\u00e9montrer qu\u2019un r\u00e9seau neuronal convolutionnel (CNN) pouvait accomplir la reconstruction de la lentille 10 millions de fois plus rapidement que les algorithmes conventionnels. Cette preuve de concept arrive juste \u00e0 temps pour permettre l&#8217;analyse de la quantit\u00e9 ph\u00e9nom\u00e9nale de donn\u00e9es qui sera produite par les t\u00e9lescopes \u00e0 champs larges dans la prochaine d\u00e9cennie. Nous devrons aussi adapter des architectures comme les machines \u00e0 inf\u00e9rences r\u00e9currentes (RIM) pour automatiser le processus de reconstruction. Les besoins scientifiques de notre mission n\u00e9cessitera d&#8217;adapter l&#8217;architecture de nos mod\u00e8les pour l\u2019estimation des incertitudes.<\/p>\n<p><strong>Hatim Belgharbi<\/strong> (Polytechnique Montr\u00e9al, Jean Provost)<\/p>\n<ul>\n<li>Microscopie de localisation par ultrasons fonctionnelle 3D (fULM)<\/li>\n<\/ul>\n<p style=\"padding-left: 40px;\">L&#8217;imagerie fonctionnelle c\u00e9r\u00e9brale permet de mieux comprendre quelles r\u00e9gions du cerveau sont impliqu\u00e9es dans diff\u00e9rents types de t\u00e2ches. Il est possible de r\u00e9aliser ce type d&#8217;analyse \u00e0 l&#8217;aide, par exemple, de l&#8217;imagerie par r\u00e9sonance magn\u00e9tique, mais \u00e0 une r\u00e9solution spatiotemporelle limit\u00e9e (de l\u2019ordre du millim\u00e8tre et de la seconde). Plus r\u00e9cemment, une autre technique, la microscopie de localisation 2D a permis de drastiquement augmenter la r\u00e9solution spatiale des ultrasons (5 milli\u00e8mes de millim\u00e8tre), mais puisqu&#8217;elle requiert la d\u00e9tection de microbulles inject\u00e9es individuelles (approuv\u00e9es en clinique), sa r\u00e9solution temporelle \u00e9tait insuffisante pour d\u00e9tecter l&#8217;activation du cerveau (dans l\u2019ordre des minutes). Le laboratoire de Jean Provost a r\u00e9cemment d\u00e9velopp\u00e9 une nouvelle technique d&#8217;imagerie appel\u00e9e Microscopie de Localisation Ultrasonore Dynamique 3D (dMLU-3D), qui permet d&#8217;atteindre la m\u00eame r\u00e9solution spatiale en trois dimensions plut\u00f4t que deux et aussi une r\u00e9solution \u00e9lev\u00e9e pour les ph\u00e9nom\u00e8nes p\u00e9riodiques (de l\u2019ordre de la milliseconde). La technique permet la visualisation de la microvasculature c\u00e9r\u00e9brale (morphologie), mais la visualisation de l\u2019activit\u00e9 c\u00e9r\u00e9brale n\u2019a pas encore \u00e9t\u00e9 d\u00e9velopp\u00e9e (fonction). La mod\u00e9lisation de ce qui caract\u00e9rise une activation c\u00e9r\u00e9brale d\u00e9pend de plusieurs param\u00e8tres non lin\u00e9aires dont il n\u2019existe pas de v\u00e9rit\u00e9 terrain \u00e0 l\u2019\u00e9chelle de la microvasculature in-vivo, alors l\u2019utilisation d\u2019un r\u00e9seau de neurones convolutionnel (CNN) s\u2019av\u00e8re pertinente \u00e0 cette application. Ce projet vise \u00e0 montrer qu&#8217;il est possible de faire de l&#8217;imagerie fonctionnelle (d\u00e9tecter l\u2019activit\u00e9 ou le manque d\u2019activit\u00e9 c\u00e9r\u00e9brale) dans tout le cerveau de rongeur \u00e0 l&#8217;aide de l&#8217;approche dMLU-3D avec une r\u00e9solution spatiotemporelle encore jamais atteinte avec d&#8217;autres m\u00e9thodes comparables. Des exp\u00e9riences seront r\u00e9alis\u00e9es afin de r\u00e9v\u00e9ler et de corr\u00e9ler l\u2019activit\u00e9 des r\u00e9gions visuelles thalamiques et corticales du cerveau du mod\u00e8le murin suivant la pr\u00e9sentation de stimuli visuels. Par la suite, ces r\u00e9sultats seront compar\u00e9s avec ceux obtenus chez des mod\u00e8les animaux de la schizophr\u00e9nie (d\u00e9veloppemental, pharmacologique, l\u00e9sionnel ou g\u00e9n\u00e9tique) afin de v\u00e9rifier l&#8217;hypoth\u00e8se que ce d\u00e9sordre est caract\u00e9ris\u00e9 par une alt\u00e9ration des connexions entre le cortex visuel et le thalamus. Ce projet serait la toute premi\u00e8re d\u00e9monstration de la faisabilit\u00e9 de l&#8217;imagerie fonctionnelle c\u00e9r\u00e9brale par ultrasons superr\u00e9solus en 2D et en 3D, permettant la cartographie de l&#8217;activation c\u00e9r\u00e9brale de la totalit\u00e9 du cerveau de rongeur ou d\u2019autres petits animaux, tel le chat, pour des \u00e9tudes pr\u00e9-cliniques permettant \u00e0 terme de mieux comprendre certaines pathologies et menant potentiellement \u00e0 un meilleur diagnostic ou m\u00eame traitement. C\u2019est d\u2019autant plus prometteur \u00e9tant donn\u00e9 qu\u2019aucune autre modalit\u00e9 d\u2019imagerie peut atteindre une r\u00e9solution aussi fine, avec une profondeur d\u2019imagerie suffisante et ce, de mani\u00e8re non invasive.<\/p>\n<p><strong>Marie-H\u00e9l\u00e8ne Bourget<\/strong> (Polytechnique Montr\u00e9al, Julien Cohen-Adad)<\/p>\n<ul>\n<li>Segmentation automatique d\u2019images histologiques par apprentissage profond<\/li>\n<\/ul>\n<p style=\"padding-left: 40px;\">Les axones de la mati\u00e8re blanche sont le prolongement des neurones, et constituent les autoroutes du syst\u00e8me nerveux central. Une gaine lipidique, la my\u00e9line, entoure ces axones permettant la conduction plus rapide de l\u2019influx nerveux. Des maladies neurod\u00e9g\u00e9n\u00e9ratives comme la scl\u00e9rose en plaques ou encore des traumatismes menacent l\u2019int\u00e9grit\u00e9 des axones my\u00e9linis\u00e9s, ce qui peut mener \u00e0 des d\u00e9ficits sensoriels ou moteurs tels que la douleur ou la parapl\u00e9gie. Afin de d\u00e9velopper de nouveaux traitements, les chercheurs en neurosciences ont besoin de quantifier avec pr\u00e9cision la morphom\u00e9trie de ces axones (taille, \u00e9paisseur de my\u00e9line, etc.). Mon laboratoire d\u2019accueil NeuroPoly a d\u00e9velopp\u00e9 le logiciel AxonDeepSeg permettant de faire la segmentation automatique de neurones sur des images histologiques par des algorithmes d\u2019apprentissage profond. Cependant, AxonDeepSeg manque de robustesse vis-\u00e0-vis de la variabilit\u00e9 qui peut exister selon les param\u00e8tres d\u2019acquisition et la qualit\u00e9 des images ainsi que selon les esp\u00e8ces. Ce projet vise donc \u00e0 d\u00e9velopper des mod\u00e8les robustes de segmentation de neurones par l\u2019adaptation et l\u2019impl\u00e9mentation de m\u00e9thodes innovantes de segmentation par apprentissage profond (Adaptation de domaine, MixUp, FiLM). Le potentiel de g\u00e9n\u00e9ralisation des algorithmes d\u00e9velopp\u00e9s sera valid\u00e9 \u00e0 l\u2019aide de bases de donn\u00e9es de microscopie incluant diverses modalit\u00e9s d\u2019imagerie (optique, \u00e9lectronique \u00e0 balayage, \u00e9lectronique en transmission), esp\u00e8ces, organes et pathologies. De plus, les mod\u00e8les d\u00e9velopp\u00e9s et les donn\u00e9es g\u00e9n\u00e9r\u00e9es seront rendus publics en acc\u00e8s libre et document\u00e9s afin de permettre \u00e0 de nombreux chercheurs et cliniciens en neurosciences de les utiliser. Cet outil permettra \u00e9galement de faire la validation d\u2019autres modalit\u00e9s d\u2019imagerie essentielles dans la recherche sur les maladies neurod\u00e9g\u00e9n\u00e9ratives comme l\u2019imagerie par r\u00e9sonance magn\u00e9tique quantitative non-invasive, et ainsi augmenter la quantit\u00e9 de donn\u00e9es utilisables par les chercheurs.<\/p>\n<p><strong>Jo\u00eblle Cormier<\/strong> (HEC Montr\u00e9al, Val\u00e9rie B\u00e9langer)<\/p>\n<ul>\n<li>Analyse du transport d\u2019urgence a\u00e9rien dans les r\u00e9gions \u00e9loign\u00e9es du Qu\u00e9bec<\/li>\n<\/ul>\n<p style=\"padding-left: 40px;\">Dans un objectif d\u2019offrir des soins sp\u00e9cialis\u00e9s \u00e0 l\u2019ensemble de sa population, le Qu\u00e9bec peut compter sur le programme d\u2019\u00c9vacuation aerom\u00e9dicales du Qu\u00e9bec (EVAQ) mis en place par le gouvernement. L\u2019offre de service permet de transf\u00e9rer des patients depuis les diff\u00e9rentes r\u00e9gions du Qu\u00e9bec vers des centres sp\u00e9cialis\u00e9s de Qu\u00e9bec et Montr\u00e9al afin de leur offrir les soins n\u00e9cessaires, le tout entour\u00e9 d\u2019une \u00e9quipe m\u00e9dicale adapt\u00e9e \u00e0 leur condition et leur niveau d\u2019urgence. Plusieurs des services offerts par l\u2019EVAQ ont connu une augmentation de la demande durant la derni\u00e8re d\u00e9cennie. La pr\u00e9sente recherche vise \u00e0 b\u00e2tir un outil de simulation qui permettra de simuler diff\u00e9rentes utilisations des ressources. L\u2019analyse des diff\u00e9rents sc\u00e9narios permettra de faire des recommandations \u00e0 l\u2019\u00c9VAQ sur les actions \u00e0 prendre afin d\u2019offrir le meilleur niveau de service possible aux populations des r\u00e9gions. Il y a beaucoup \u00e0 apprendre sur le mod\u00e8le instaur\u00e9 au Qu\u00e9bec, tant au niveau de la planification strat\u00e9gique des appareils et des trajets, qu\u2019au niveau de la coordination et des op\u00e9rations au quotidien. La densit\u00e9 de population, les distances \u00e0 franchir et les conditions m\u00e9t\u00e9orologiques difficiles sont des facteurs d\u00e9terminants \u00e0 consid\u00e9rer dans leur unicit\u00e9.<\/p>\n<p><strong>Edward Hall\u00e9-Hannan<\/strong> (Polytechnique Montr\u00e9al, S\u00e9bastien Le Digabel)<\/p>\n<ul>\n<li>Optimisation de l\u2019entra\u00eenement des r\u00e9seaux de neurones profonds \u00e0 partir d\u2019extensions de l\u2019algorithme MADS sur les hyperparam\u00e8tres de type variable de cat\u00e9gorie<\/li>\n<\/ul>\n<p style=\"padding-left: 40px;\">Ce projet de ma\u00eetrise vise \u00e0 optimiser l\u2019entra\u00eenement des r\u00e9seaux de neurones profonds \u00e0 partir d\u2019extensions de l\u2019algorithme MADS sur les hyperparam\u00e8tres de type variable de cat\u00e9gorie. Ces hyperparam\u00e8tres sont g\u00e9n\u00e9ralement choisis de mani\u00e8re arbitraire ou heuristique. Or, la plupart des algorithmes d\u2019optimisation d\u00e9velopp\u00e9s solutionnent des probl\u00e8mes o\u00f9 les variables sont de type continu ou entier. En d\u2019autres mots, il existe peu de m\u00e9thodes d\u2019optimisation pouvant traiter efficacement les variables de cat\u00e9gorie. Cependant, puisque ces variables sont discr\u00e8tes, il est possible de construire et d\u2019explorer un espace de variables discr\u00e9tis\u00e9es avec les m\u00e9thodes d\u2019optimisation dites recherche directe. Le projet de recherche a pour objectif d\u2019adapter les r\u00e9cents d\u00e9veloppements de l\u2019algorithme MADS (\u00ab&nbsp;Mesh Adaptive Direct Search&nbsp;\u00bb) aux variables de cat\u00e9gorie, notamment pour le traitement des contraintes et l\u2019int\u00e9gration d\u2019un treillis anisotrope dynamique. Plus pr\u00e9cis\u00e9ment, nous nous int\u00e9ressons \u00e0 optimiser plus rigoureusement les hyperparam\u00e8tres des r\u00e9seaux de neurones profonds, afin d\u2019entra\u00eener plus intelligemment les mod\u00e8les d\u2019intelligence artificielle. Plus particuli\u00e8rement, les hyperparam\u00e8tres \u00e9tudi\u00e9s seront&nbsp;: la fonction de perte ; les extensions et les modifications \u00e0 l\u2019algorithme de r\u00e9tropropagation (ADAM, RMSProp, etc.) ainsi que les r\u00e9gulateurs (LASSO, \u00ab&nbsp;Ridge regression&nbsp;\u00bb, etc.). Les m\u00e9canismes d\u00e9velopp\u00e9s pourront \u00e9galement servir \u00e0 mod\u00e9liser la topologie des r\u00e9seaux (nombres de couches, nombres de neurones, etc.) En effet, dans le cadre de l\u2019algorithme MADS, le traitement des variables de cat\u00e9gorie pourraient s\u2019\u00e9tendre \u00e0 des variables discr\u00e8tes, dont la valeur modifie la dimension du probl\u00e8me. En pratique, le syst\u00e8me r\u00e9sultant permettra donc, pour la premi\u00e8re fois, d\u2019optimiser simultan\u00e9ment les hyperparam\u00e8tres reli\u00e9s \u00e0 l\u2019entra\u00eenement et ceux reli\u00e9s \u00e0 la topologie.<\/p>\n<p><strong>Dongyan Lin<\/strong> (McGill University, Blake Richards)<\/p>\n<ul>\n<li>Analyzing mouse hippocampal &#8220;time cell&#8221; activities during memory task with machine learning approaches<\/li>\n<\/ul>\n<p style=\"padding-left: 40px;\">Previous studies have identified hippocampal \u201ctime cells\u201d in CA1 that bridge the temporal gap between discontiguous events by firing in tiling patterns during the delay period of memory tasks, such as alternative maze (Pastalkova et al., 2008) and object-odor pairing tasks (MacDonald et al., 2011). However, recent findings have argued that this tiling might be an analysis artifact due to cell-sorting because it also appears in tasks with no memory load (Salz et al., 2016). To address this discrepancy, our collaborators have collected calcium recordings in mouse hippocampal CA1 region during trial unique, nonmatch-to-location (TUNL) task (Talpos et al., 2010) and showed tiling patterns. Our objective is to use computational methods to determine if these patterns are meaningful. To do this, we will first train decoders on the calcium recordings to decode sample for each trial, with temporal sequences preserved (i.e. sorted tiling columns) or shuffled (i.e. randomized columns). If the tiling patterns are indeed meaningful, we would expect to see higher accuracy of the decoder in the preserved sequences. Our next step is to construct a simulated reinforcement learning agent on simulated TUNL task to see whether there exists a consistent tiling pattern in the activities of the neural networks of the reinforcement learning agent. If so, it would suggest that these patterns play a role in preserving information about the sample location during the delay period as a solution to the task. If not, it would suggest that the tiling patterns previously observed in memory tasks could merely be a ubiquitous artifact. Our findings would have a significant impact on the current view of hippocampal \u201ctime cells\u201d as well as the functional segregation of the brain.<\/p>\n<p><strong>Yiqun (Arlene) Lu<\/strong> (Polytechnique Montr\u00e9al, Guillaume-Alexandre Bilodeau)<\/p>\n<ul>\n<li>Jumpy, Hierarchical and Adversarial Variational Video Prediction<\/li>\n<\/ul>\n<p style=\"padding-left: 40px;\">This project is in the context of intelligent transportation systems. To improve road user detection and tracking, we want to predict their position in future frames using video prediction. However, predicting high fidelity videos over long time scale is notoriously difficult. Current video prediction models either diverges from real samples after a few frames or fail to capture stochasticity in the videos, resulting in bad prediction performance for long videos. In order to overcome this difficulty, new models with ability to do jumpy or hierarchical video prediction are proposed by the AI community. In this proposal, we propose to further develop these ideas and explore new models for stochastic video prediction that is able to do jumpy predictions in a hierarchical manner. We mainly want to explore two research problems: (1) How to do stochastic jumpy video predictions. (2) How to combine jumpy prediction with temporal abstraction.<\/p>\n<p><strong>Andrei Lupu<\/strong> (McGill University, Doina Precup)<\/p>\n<ul>\n<li>Emergent Behaviour in Multi-Agent Reinforcement Learning<\/li>\n<\/ul>\n<p style=\"padding-left: 40px;\">This project aims for the investigation of intricate emergent behaviours in large scale multi-agent reinforcement learning (MARL). Of particular concern are the behaviours of agents in settings where they are tightly interdependent to the point of nearly composing a single entity. Such settings will draw strong inspiration from biological systems, and be achieved either through a shared common reward or through complex and necessary interactions. Because large interconnected populations of agents present a novel collection of settings complete with new challenges, this project will force a rethinking of well-established reinforcement learning practices, all while probing the limits of their scalability. Furthermore, enabling MARL systems that simultaneously achieve large population scales and appropriate complexity will allow for better modelling of intricate phenomena that have been out of reach of previous artificial intelligence methods. This would potentially result in far-reaching benefits in other scientific disciplines, thus broadening the range of applications of reinforcement learning and simultaneously opening it to easier idea cross-pollination from other fields. These settings will be studied empirically by analyzing the behaviour of existing MARL algorithms, and by comparing and contrasting them to new approaches that allow for more complex interactions between agents. The analysis of the results will be performed quantitatively on the basis of standard reinforcement learning and game theoretic methodology, and qualitatively in light of the principles of behavioural biology. The implementation of the environments and the MARL models will be done with modularity and concurrency in mind and the code-base will then be openly released.<\/p>\n<p><strong>Nicholas Meade<\/strong> (McGill University, Siva Reddy)<\/p>\n<ul>\n<li>Stylistic Controls for Neural Text Generation<\/li>\n<\/ul>\n<p style=\"padding-left: 40px;\">Deep learning-based approaches to text generation have proven effective in recent years, with many models able to generate realistic text, often exhibiting higher-order structure. While these models produce high-quality samples, there is usually little control provided over what is specifically generated. Recently, work has begun in this area, but much remains to be explored. This application proposes research towards controllable text generation by implementing a variety of stylistic controls that can be used to influence what is sampled from a neural language model. In my previous work, we developed a conditional generative model for music. We demonstrated that we could control for a variety of characteristics during generation by providing the model with an additional externally-specified input called the control signal. For instance, in this work, we trained a model using a composer-based control signal. This signal identified the composer of each piece on which the model was trained. After training, we used the control signal to produce samples of music in the style of specific composers, for instance, Bach and Beethoven. Based on my previous work with music, we are now interested in implementing a similar set of controls for generating text. Such a set of stylistic controls would extend the practical utility of text generated from neural language models. We plan to explore generation methods involving supervised controls and latent (disentangled) controls.<\/p>\n<p><strong>Marie-Eve Picard<\/strong> (Universit\u00e9 de Montr\u00e9al, Pierre Rainville)<\/p>\n<ul>\n<li>Utilisation d&#8217;approches d&#8217;apprentissage automatique pour l&#8217;identification d&#8217;une signature c\u00e9r\u00e9brale de l&#8217;expression faciale de la douleur<\/li>\n<\/ul>\n<p style=\"padding-left: 40px;\">L&#8217;expression faciale est un outil important pour communiquer diverses informations, notamment la manifestation d&#8217;un \u00e9tat de douleur, la pr\u00e9sence d&#8217;une menace imm\u00e9diate dans l&#8217;environnement et un \u00e9ventuel besoin d&#8217;aide. Les dimensions sensorielle (intensit\u00e9) et affective (caract\u00e8re d\u00e9plaisant) de la douleur peuvent \u00eatre encod\u00e9es dans les mouvements faciaux. Les techniques d\u2019analyse jusqu\u2019\u00e0 pr\u00e9sent utilis\u00e9es pour examiner la relation entre l\u2019expression faciale et l\u2019activit\u00e9 c\u00e9r\u00e9brale lors de l\u2019exp\u00e9rience de la douleur poss\u00e8dent plusieurs limitations statistiques par rapport \u00e0 l\u2019\u00e9valuation de l\u2019activit\u00e9 c\u00e9r\u00e9brale spatialement distribu\u00e9e. L&#8217;objectif principal du projet propos\u00e9 est de mieux comprendre les m\u00e9canismes neuronaux qui sous-tendent l&#8217;expression faciale de la douleur. Des donn\u00e9es d&#8217;imagerie par r\u00e9sonance magn\u00e9tique fonctionnelle (IRMf) seront utilis\u00e9es pour analyser les changements dans l&#8217;activit\u00e9 c\u00e9r\u00e9brale en r\u00e9ponse \u00e0 des stimuli douloureux (mais non dommageables). Plus sp\u00e9cifiquement, ce projet vise \u00e0 utiliser des approches d&#8217;apprentissage automatique (c&#8217;est-\u00e0-dire l&#8217;analyse de mod\u00e8les multivari\u00e9s) pour d\u00e9velopper une signature c\u00e9r\u00e9brale de l&#8217;expression faciale de la douleur afin de pr\u00e9dire les changements faciaux en r\u00e9ponse \u00e0 des stimuli douloureux dans diff\u00e9rents contextes&nbsp;: douleur phasique (stimulation courte), douleur tonique (stimulation longue), et modulation des dimensions sensorielle et affective de la douleur. En bref, ce projet permettra de r\u00e9soudre certaines lacunes des analyses univari\u00e9s pr\u00e9c\u00e9demment utilis\u00e9es afin de d\u00e9terminer avec une meilleure pr\u00e9cision les bases neurales de l\u2019expression faciale de la douleur et de faire progresser de mani\u00e8re significative notre compr\u00e9hension des m\u00e9canismes c\u00e9r\u00e9braux qui sous-tendent la communication non verbale.<\/p>\n<p><strong>Myriam Prasow-\u00c9mond<\/strong> (Universit\u00e9 de Montr\u00e9al, Julie Hlavacek-Larrondo)<\/p>\n<ul>\n<li>Les premi\u00e8res images d&#8217;exoplan\u00e8tes orbitant autour de naines blanches, d&#8217;\u00e9toiles \u00e0 neutrons et de trous noirs<\/li>\n<\/ul>\n<p style=\"padding-left: 40px;\">Les binaires X, form\u00e9s d&#8217;une \u00e9toile orbitant autour d&#8217;un objet compact stellaire compact (naine blanche, \u00e9toile \u00e0 neutrons ou trou noir), sont des laboratoires fantastiques pour comprendre la physique dans des conditions extr\u00eames. Au cours des derni\u00e8res d\u00e9cennies, les binaires X ont fait l&#8217;objet d&#8217;une multitude d&#8217;\u00e9tudes dans diverses longueurs d&#8217;onde, conduisant \u00e0 des avanc\u00e9es remarquables dans le domaine de la physique de l&#8217;accr\u00e9tion, ainsi que dans la compr\u00e9hension de la formation de jets de particules relativistes dans de puissants champs magn\u00e9tiques. Les binaires X sont aussi d&#8217;excellents laboratoires pour comprendre les explosions de type supernova ainsi que l&#8217;effet de ces explosions sur le syst\u00e8me et son environnement. En effet, la pr\u00e9sence d&#8217;une \u00e9toile \u00e0 neutrons ou d&#8217;un trou noir dans ces syst\u00e8mes implique directement que l&#8217;\u00e9toile (et ses potentielles plan\u00e8tes) survivent \u00e0 ces explosions. Plusieurs \u00e9tudes montrent que les plan\u00e8tes et les naines brunes peuvent exister dans une multitude d&#8217;environnements, tels que celles qui orbitent tr\u00e8s proche de leur \u00e9toile h\u00f4te (Jupiters chaudes) ou celles qui orbitent \u00e0 des distances de centaines d&#8217;unit\u00e9s astronomiques de l&#8217;\u00e9toile. Ces d\u00e9couvertes montrent que la formation et la survie des plan\u00e8tes sont mal comprises. Par cons\u00e9quent, ce projet am\u00e8ne un nouveau point de vue, soit celui des conditions extr\u00eames. Bref, on pourra \u00e9tudier plusieurs binaires X et des donn\u00e9es des t\u00e9lescopes NIRC2\/KECK (visible) et NOEMA (millim\u00e9trique) ont d\u00e9j\u00e0 \u00e9t\u00e9 acquises en 2018, et d&#8217;autres demandes de temps sont en cours. Selon une analyse pr\u00e9liminaire, la pr\u00e9sence d&#8217;objets astrophysiques est confirm\u00e9e, et donc ce projet garantit des r\u00e9sultats surprenants pour la communaut\u00e9 de l&#8217;astrophysique.<\/p>\n<p><strong>Chence Shi<\/strong> (HEC Montr\u00e9al, Jian Tang)<\/p>\n<ul>\n<li>Addressing the retrosynthesis problem using a graph-to-graph translation network<\/li>\n<\/ul>\n<p style=\"padding-left: 40px;\">Retrosynthesis analysis, which aims to identify a set of reactant graphs to synthesize a target molecule, is a fundamental problem in computational chemistry and is of central importance to the organic synthesis planing as well as drug discovery. The problem is challenging as the search space of all possible transformations is very huge. For decades, people have been seeking to assist chemists in retrosynthesis analysis with modern computing algorithms. Most existing machine learning works on this task rely on reaction templates that define the subgraph patterns of a set of chemical reactions, which require expensive graph isomorphism and suffer from poor generalization on unseen molecule structures.<\/p>\n<p style=\"padding-left: 40px;\">To address the above limitations, in this project, we formulate the retrosynthesis prediction as a graph-to-graph translation task, i.e., translating a product graph to a set of reactant graphs, and propose a novel template-free approach to tackle the problem. We will show that our method excludes the need of domain knowledge, and scales well to large datasets. We will also empirically verify the superiority of our method on the benchmark data set.<\/p>\n<p><strong>Shi Tianyu<\/strong> (McGill University, Luis Miranda-Moreno)<\/p>\n<ul>\n<li>A Multi-agent Decision and Control Framework for Mixed-autonomy Transportation System<\/li>\n<\/ul>\n<p style=\"padding-left: 40px;\">As the autonomous vehicle becomes more and more popular. Recently, there has been a new emphasis on traffic control in the context of mixed-autonomy, where only a fraction of vehicles are connected autonomous vehicles and interacting with human-driven vehicles. As in a mixed autonomy system, there are several challenges. The first challenge is how to encourage different agents&#8217; cooperation so as to maximize the total returns of the whole system. For example, when there is a gap in front of the adjacent line of the autonomous vehicle, if the autonomous vehicle cuts in immediately, the surrounding vehicle in the adjacent line will also decrease its speed sharply, which will end up a shock wave in traffic flow. Instead, if the autonomous vehicle learns to cooperate with other agents, it will adjust its speed steadily and try to mitigate the negative impact on the whole system. The second challenge is how to improve the communication efficiency in multi-agent system. As autonomous vehicle has different characteristics with human-driven agent, for example, their reacting time and action may be different. Therefore, how to formalize personalized policy for each agent is also worth to explore. The third challenge is how to explore expert knowledge (e.g. green wave, max pressure, actuated control) in transportation domain to improve the training efficiency and performance. Our overall goal of this project is to design effective decision and control framework for an efficient and safe mixed autonomy system by mitigating the shockwave and improving the transportation efficiency. To address the aforementioned problems, we will develop a novel multi agent decision framework based on deep reinforcement learning to improve the decision making and control performance of the agents in mixed autonomy system.<\/p>\n<p><strong>Rey Wiyatno<\/strong> (Universit\u00e9 de Montr\u00e9al, Liam Paull)<\/p>\n<ul>\n<li>Exploiting Experiences and Priors in Semantic Visual Navigation<\/li>\n<\/ul>\n<p style=\"padding-left: 40px;\">Robotics has always been anticipated to revolutionize the world. However, despite the significant progress over the past few decades, robots have yet to be able to reliably navigate within an unstructured indoor environment. Semantic visual navigation is the task of navigating within a possibly unknown environment using only visual sensors, such as asking a household robot agent to \u201cgo to the kitchen\u201d. Traditional \u201cmodular\u201d methods combine a Simultaneous Localization and Mapping (SLAM) component with separate search, planning, and control modules. However, these methods do not scale well to large environments, and require significant engineering efforts. Alternatively, end-to-end \u201clearning\u201d solutions produce agent policies that directly infer actions from camera frames, by applying Deep Reinforcement Learning (DRL) techniques on large-scale datasets. Nevertheless, these policies tend to be reactive, do not explicitly exploit scene geometry, and are not data efficient. Furthermore, both modular and learning-based approaches do not sufficiently exploit knowledge from past task instances to improve subsequent search performance in both repeated environments as well as unseen yet similar environments. Our project explores the learning and use of spatial-semantic priors for more efficient semantic visual navigation. We aim to devise a framework that learns, updates, and exploits a topological-semantic map between discovered locations and objects within. We hypothesize that these advances will result in agents that generalize better to unseen similar environments, as well as becoming increasingly more efficient during repeated search queries within the same environment.<\/p>\n<p><strong>Chengyuan Zhang<\/strong> (McGill University, Lijun Sun)<\/p>\n<ul>\n<li>Statistical Modeling Framework to Understand Dynamic Traffic Patterns from Video Data<\/li>\n<\/ul>\n<p style=\"padding-left: 40px;\">Video-based traffic monitoring systems, as the backbone of modern Intelligent Transportation Systems (ITS), is playing an essential role in sensing traffic conditions and detecting abnormal events\/incidents. Semantically understanding traffic scenes and automatically mining the traffic patterns from video data of a static camera can help with traffic situation analysis and anomaly events warning. Given a video of a dynamic traffic scene with several different behaviors happening simultaneously, we want the ITS to learn and understand: \u201cHow many typical traffic patterns are in the video? How to semantically interpret these patterns? What are the rules governing the transitions between these patterns?\u201dIn this project, we will mainly focus on traffic patterns recognition and anomaly detection from video data, we will: (i) construct representation learning model to extract efficient features; and (ii) develop an unsupervised learning framework based on Bayesian nonparametrics to automatically learn the traffic patterns.<\/p>\n<\/div><\/div><div class=\"clear\"><\/div><\/div><\/div>\n<p>&nbsp;<\/p>\n","protected":false},"excerpt":{"rendered":"<p>Next planned opening: Fall 2020 IVADO excellence scholarship program for Msc IVADO\u2019s commitment to equity, diversity and inclusion and note to applicants To ensure all members of society draw equal&#8230; <\/p>\n<div class=\"clear\"><\/div>\n<p><a href=\"https:\/\/vieux.ivado.ca\/en\/ivado-scholarships\/excellence-scholarships-msc\/\" class=\"gdlr-button with-border excerpt-read-more\">Read More<\/a><\/p>\n","protected":false},"author":33,"featured_media":0,"parent":9026,"menu_order":0,"comment_status":"closed","ping_status":"closed","template":"","meta":{"footnotes":""},"acf":[],"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v22.7 - https:\/\/yoast.com\/wordpress\/plugins\/seo\/ -->\n<title>Excellence Scholarships - Msc - IVADO<\/title>\n<meta name=\"robots\" content=\"noindex, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<meta property=\"og:locale\" content=\"en_US\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Excellence Scholarships - Msc - IVADO\" \/>\n<meta property=\"og:description\" content=\"Next planned opening: Fall 2020 IVADO excellence scholarship program for Msc IVADO\u2019s commitment to equity, diversity and inclusion and note to applicants To ensure all members of society draw equal... Read More\" \/>\n<meta property=\"og:url\" content=\"https:\/\/vieux.ivado.ca\/en\/ivado-scholarships\/excellence-scholarships-msc\/\" \/>\n<meta property=\"og:site_name\" content=\"IVADO\" \/>\n<meta property=\"article:modified_time\" content=\"2020-06-22T17:33:06+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/vieux.ivado.ca\/wp-content\/uploads\/2017\/10\/BanniereExcellenceScholarships.png\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"Est. reading time\" \/>\n\t<meta name=\"twitter:data1\" content=\"46 minutes\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":\"WebPage\",\"@id\":\"https:\/\/vieux.ivado.ca\/en\/ivado-scholarships\/excellence-scholarships-msc\/\",\"url\":\"https:\/\/vieux.ivado.ca\/en\/ivado-scholarships\/excellence-scholarships-msc\/\",\"name\":\"Excellence Scholarships - Msc - IVADO\",\"isPartOf\":{\"@id\":\"https:\/\/vieux.ivado.ca\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\/\/vieux.ivado.ca\/en\/ivado-scholarships\/excellence-scholarships-msc\/#primaryimage\"},\"image\":{\"@id\":\"https:\/\/vieux.ivado.ca\/en\/ivado-scholarships\/excellence-scholarships-msc\/#primaryimage\"},\"thumbnailUrl\":\"https:\/\/vieux.ivado.ca\/wp-content\/uploads\/2017\/10\/BanniereExcellenceScholarships.png\",\"datePublished\":\"2018-12-27T18:27:28+00:00\",\"dateModified\":\"2020-06-22T17:33:06+00:00\",\"breadcrumb\":{\"@id\":\"https:\/\/vieux.ivado.ca\/en\/ivado-scholarships\/excellence-scholarships-msc\/#breadcrumb\"},\"inLanguage\":\"en-US\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\/\/vieux.ivado.ca\/en\/ivado-scholarships\/excellence-scholarships-msc\/\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\/\/vieux.ivado.ca\/en\/ivado-scholarships\/excellence-scholarships-msc\/#primaryimage\",\"url\":\"https:\/\/vieux.ivado.ca\/wp-content\/uploads\/2017\/10\/BanniereExcellenceScholarships.png\",\"contentUrl\":\"https:\/\/vieux.ivado.ca\/wp-content\/uploads\/2017\/10\/BanniereExcellenceScholarships.png\"},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\/\/vieux.ivado.ca\/en\/ivado-scholarships\/excellence-scholarships-msc\/#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Accueil\",\"item\":\"https:\/\/vieux.ivado.ca\/en\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"IVADO scholarships\",\"item\":\"https:\/\/vieux.ivado.ca\/en\/ivado-scholarships\/\"},{\"@type\":\"ListItem\",\"position\":3,\"name\":\"Excellence Scholarships &#8211; Msc\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\/\/vieux.ivado.ca\/#website\",\"url\":\"https:\/\/vieux.ivado.ca\/\",\"name\":\"IVADO\",\"description\":\"Institut de valorisation des donn\u00e9es\",\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\/\/vieux.ivado.ca\/?s={search_term_string}\"},\"query-input\":\"required name=search_term_string\"}],\"inLanguage\":\"en-US\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"Excellence Scholarships - Msc - IVADO","robots":{"index":"noindex","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"og_locale":"en_US","og_type":"article","og_title":"Excellence Scholarships - Msc - IVADO","og_description":"Next planned opening: Fall 2020 IVADO excellence scholarship program for Msc IVADO\u2019s commitment to equity, diversity and inclusion and note to applicants To ensure all members of society draw equal... Read More","og_url":"https:\/\/vieux.ivado.ca\/en\/ivado-scholarships\/excellence-scholarships-msc\/","og_site_name":"IVADO","article_modified_time":"2020-06-22T17:33:06+00:00","og_image":[{"url":"https:\/\/vieux.ivado.ca\/wp-content\/uploads\/2017\/10\/BanniereExcellenceScholarships.png"}],"twitter_card":"summary_large_image","twitter_misc":{"Est. reading time":"46 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"WebPage","@id":"https:\/\/vieux.ivado.ca\/en\/ivado-scholarships\/excellence-scholarships-msc\/","url":"https:\/\/vieux.ivado.ca\/en\/ivado-scholarships\/excellence-scholarships-msc\/","name":"Excellence Scholarships - Msc - IVADO","isPartOf":{"@id":"https:\/\/vieux.ivado.ca\/#website"},"primaryImageOfPage":{"@id":"https:\/\/vieux.ivado.ca\/en\/ivado-scholarships\/excellence-scholarships-msc\/#primaryimage"},"image":{"@id":"https:\/\/vieux.ivado.ca\/en\/ivado-scholarships\/excellence-scholarships-msc\/#primaryimage"},"thumbnailUrl":"https:\/\/vieux.ivado.ca\/wp-content\/uploads\/2017\/10\/BanniereExcellenceScholarships.png","datePublished":"2018-12-27T18:27:28+00:00","dateModified":"2020-06-22T17:33:06+00:00","breadcrumb":{"@id":"https:\/\/vieux.ivado.ca\/en\/ivado-scholarships\/excellence-scholarships-msc\/#breadcrumb"},"inLanguage":"en-US","potentialAction":[{"@type":"ReadAction","target":["https:\/\/vieux.ivado.ca\/en\/ivado-scholarships\/excellence-scholarships-msc\/"]}]},{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/vieux.ivado.ca\/en\/ivado-scholarships\/excellence-scholarships-msc\/#primaryimage","url":"https:\/\/vieux.ivado.ca\/wp-content\/uploads\/2017\/10\/BanniereExcellenceScholarships.png","contentUrl":"https:\/\/vieux.ivado.ca\/wp-content\/uploads\/2017\/10\/BanniereExcellenceScholarships.png"},{"@type":"BreadcrumbList","@id":"https:\/\/vieux.ivado.ca\/en\/ivado-scholarships\/excellence-scholarships-msc\/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Accueil","item":"https:\/\/vieux.ivado.ca\/en\/"},{"@type":"ListItem","position":2,"name":"IVADO scholarships","item":"https:\/\/vieux.ivado.ca\/en\/ivado-scholarships\/"},{"@type":"ListItem","position":3,"name":"Excellence Scholarships &#8211; Msc"}]},{"@type":"WebSite","@id":"https:\/\/vieux.ivado.ca\/#website","url":"https:\/\/vieux.ivado.ca\/","name":"IVADO","description":"Institut de valorisation des donn\u00e9es","potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/vieux.ivado.ca\/?s={search_term_string}"},"query-input":"required name=search_term_string"}],"inLanguage":"en-US"}]}},"_links":{"self":[{"href":"https:\/\/vieux.ivado.ca\/en\/wp-json\/wp\/v2\/pages\/10981\/"}],"collection":[{"href":"https:\/\/vieux.ivado.ca\/en\/wp-json\/wp\/v2\/pages\/"}],"about":[{"href":"https:\/\/vieux.ivado.ca\/en\/wp-json\/wp\/v2\/types\/page\/"}],"author":[{"embeddable":true,"href":"https:\/\/vieux.ivado.ca\/en\/wp-json\/wp\/v2\/users\/33\/"}],"replies":[{"embeddable":true,"href":"https:\/\/vieux.ivado.ca\/en\/wp-json\/wp\/v2\/comments\/?post=10981"}],"version-history":[{"count":0,"href":"https:\/\/vieux.ivado.ca\/en\/wp-json\/wp\/v2\/pages\/10981\/revisions\/"}],"up":[{"embeddable":true,"href":"https:\/\/vieux.ivado.ca\/en\/wp-json\/wp\/v2\/pages\/9026\/"}],"wp:attachment":[{"href":"https:\/\/vieux.ivado.ca\/en\/wp-json\/wp\/v2\/media\/?parent=10981"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}