We are glad to announce that Canadian AI 2022 will welcome the following confirmed keynote speakers:
- Nicolas Papernot (University of Toronto and Vector Institute)
- Diana Inkpen (University of Ottawa)
- David Poole (University of British Columbia)
- Julia Stoyanovich (New York University)
University of Toronto and Vector Institute
"Is Differential Privacy a Silver Bullet for Machine Learning?"
Some machine learning applications involve training data that is sensitive, such as the medical histories of patients in a clinical trial. A model may inadvertently and implicitly store some of its training data; careful analysis of the model may therefore reveal sensitive information. To address this problem, algorithms for private machine learning have been proposed. In this talk, we first show that training neural networks with rigorous privacy guarantees like differential privacy requires rethinking their architectures with the goals of privacy-preserving gradient descent in mind. Second, we explore how private aggregation surfaces the synergies between privacy and generalization in machine learning. Third, we present recent work towards a form of collaborative machine learning that is both privacy-preserving in the sense of differential privacy, and confidentiality-preserving in the sense of the cryptographic community. We motivate the need for this new approach by showing how existing paradigms like federated learning fail to preserve privacy in these settings.
Biography: Nicolas Papernot is an Assistant Professor in the Department of Electrical and Computer Engineering and the Department of Computer Science at the University of Toronto. He is also a faculty member at the Vector Institute where he holds a Canada CIFAR AI Chair, and a faculty affiliate at the Schwartz Reisman Institute. His research interests span the security and privacy of machine learning. Nicolas is a Connaught Researcher and was previously a Google PhD Fellow. His work on differentially private machine learning received a best paper award at ICLR 2017. He is an associate chair of IEEE S&P (Oakland) and an area chair of NeurIPS. He earned his Ph.D. at the Pennsylvania State University, working with Prof. Patrick McDaniel. Upon graduating, he spent a year as a research scientist at Google Brain where he still spends some of his time.
University of Ottawa
"Estimating Depression and Suicide Ideation Levels from Social Media for the Canadian Population"
Millions of people in Canada and elsewhere suffer from mental illness, and not all receive adequate treatment. Identifying people with mental illness requires initiation from those in need and available medical services. These resources might not always be available. The common practice is to rely on clinical data, which is generally collected after the illness is developed and reported. Moreover, such clinical data is incomplete and hard to obtain. An alternative data source is to conduct surveys through phone calls, interviews, or mail, but this is costly and time-consuming. Social media analysis has brought advances in leveraging population data to understand mental health problems. Thus, analyzing social media posts can be an essential alternative for identifying mental disorders at message level and user level, then scale it at population level. In this research, we explore the task of automatically analysing social media textual data using Natural Language Processing and Machine Learning techniques to detect signs of mental health disorders, in particular depression and suicide ideation. Considering the lack of comprehensive annotated data in this field, we propose a methodology for transfer learning to exploit the information contained in a training sample of social media data and leverage it on different test datasets, in order to choose the best-generalized models. In our proposed models, we use feature engineering with supervised machine learning algorithms (such as SVM, LR, RF, XGBoost, and GBDT), as well as deep learning algorithms (such as LSTM, Bi-LSTM, and CNNs). We choose the best model for depression detection and the best model for suicide ideation detection, on the user-level test data. Then, these two models (that are based on CNN) are used to produce estimations on the Canadian population sample data.
Biography: Diana Inkpen is a Professor at the University of Ottawa, in the School of Electrical Engineering and Computer Science. She obtained her Ph.D. from the University of Toronto, Department of Computer Science. She has a M.Sc. and B.Eng. degree in Computer Science and Engineering from the Technical University of Cluj-Napoca, Romania. Her research is in applications of Natural Language Processing and Text Mining. She organized seven international workshops and she was a program co-chair for the 25th Canadian Conference on Artificial Intelligence (AI 2012, Toronto, ON, May 2012) conference. She is the editor-in-chief of the Computational Intelligence journal and the associate editor for the Natural Language Engineering journal. She has published a book on Natural Language Processing for Social Media (Morgan and Claypool Publishers, Synthesis Lectures on Human Language Technologies, the third edition appeared in 2020), 10 book chapters, more than 35 journal articles, and more than 125 conference papers. She received many research grants, from which the majority include intensive industrial collaborations.
University of British Columbia
"Learning and reasoning about entities and relations under uncertainty: a story of romance and disappointment"
Over the decades researchers have fallen in love with many technologies for learning and reasoning about entities and relations under uncertainty, from probabilistic logic programs to Markov logic to knowledge graphs to embedding-based models to graph neural networks. Unfortunately, none of them are the silver bullet we had hoped for. I will talk about the foundations of each of these technologies, what it is good for and why it isn't, by itself, the solution. I will argue that we need get back to foundations. We need to be suspicious of prediction engines trained on data for arbitrary subjects with no models of how the data was generated (including why some information is missing), models that assume big data, even though there is little data about almost every entity, measures of success that don't correspond to the reasons we may want to learn, and overfitting on test sets that were designed to show off the abilities of previous models. Although, or maybe because, there are no magic bullets, I will explain why this is an exciting area to do science in.
Biography: David Poole is a Professor of Computer Science at the University of British Columbia. He is known for his work on combining logic and probability, probabilistic inference, relational probabilistic models, statistical relational AI and semantic science. He is the co-author of an upcoming 3rd edition of an introductory AI textbook (Cambridge University Press, 2010, 2nd edition 2017, and 3rd edition 2023), co-author of “Statistical Relational Artificial Intelligence: Logic, Probability, and Computation", (Morgan & Claypool 2016), co-editor of "Introduction to Lifted Inference" (MIT Press 2021), and co-author of an earlier AI textbook (Oxford University Press, 1998). He is a former chair of the Association for Uncertainty in Artificial Intelligence, the winner of the Canadian AI Association (CAIAC) 2013 Lifetime Achievement Award, and is a Fellow of the Association for the Advancement Artificial Intelligence (AAAI) and CAIAC. See http://www.cs.ubc.ca/~poole/publications.html for a list of his publications.
New York University
"Building Data Equity Systems"
Equity as a social concept — treating people differently depending on their endowments and needs to provide equality of outcome rather than equality of treatment — lends a unifying vision for ongoing work to operationalize ethical considerations across technology, law, and society. In my talk I will present a vision for designing, developing, deploying, and overseeing data-intensive systems that consider equity as an essential objective. I will discuss ongoing technical work, and will place this work into the broader context of policy, education, and public outreach.
Biography: Julia Stoyanovich is an Institute Associate Professor of Computer Science & Engineering at the Tandon School of Engineering, Associate Professor of Data Science at the Center for Data Science, and Director of the Center for Responsible AI at New York University (NYU). Her research focuses on responsible data management and analysis: on operationalizing fairness, diversity, transparency, and data protection in all stages of the data science lifecycle. She established the "Data, Responsibly" consortium and served on the New York City Automated Decision Systems Task Force, by appointment from Mayor de Blasio. Julia developed and has been teaching courses on Responsible Data Science at NYU, and is a co-creator of an award-winning comic book series on this topic. In addition to data ethics, Julia works on the management and analysis of preference and voting data, and on querying large evolving graphs. She holds M.S. and Ph.D. degrees in Computer Science from Columbia University, and a B.S. in Computer Science and in Mathematics & Statistics from the University of Massachusetts at Amherst. She is a recipient of an NSF CAREER award and a Senior Member of the ACM.