Montreal Time Series Meetup - Winter 2023

We are calling all Time Series enthusiasts! Borealis AI and Moov AI present the Winter 2023 installment of our time series talks.
You're invited to our IN-PERSON meetup on January 12th from 5:30 pm - 8:00 pm ET. Join us at Moov AI’s Office for networking, Pizza, and some great talks.

You'll hear from two industry practitioners working at the forefront of Time Series research and business applications. There will be plenty of Q&A time where you can pick our speaker's brains.

Our Speakers:

Simon Dagenais, AI Research Scientist at Volta Charging

• Bio: Holding a masters in economics, Simon is now an experienced Data Scientist. As one of MoovAI's first employees, he had the chance to work on a myriad of different data science problems and also contributed to their data and model validation platform, SnitchAI. He is now launching Iuvo-AI, a AI and Data consulting firm, along with colleagues he met at Volta Charging.
• Talk: Anomaly detection in time-series: how Isolation Forests can improve Hydro-Quebec's data - During the 2022 version of the HackQC, Simon and the Iuvo-AI team proposed a solution to clean and augment Hydro-Québec's data. Amongst other methods, Isolation Forests proved useful. In this talk, Simon will explain their methodology and how to effectively implement the unsupervised learning algorithm.

Tristan Sylvain, ML Researcher at Borealis AI

• Bio: Tristan Sylvain is a Machine Learning researcher at Borealis AI, focusing on time-series forecasting for a wide range of applications. The goal of his research is to understand the fundamental principles that underpin generalization in machine learning, with a particular emphasis on self-supervised representation learning.

• Talk: Recent progress and challenges in time-series forecasting with large-scale architectures - The performance of time-series forecasting has recently been greatly improved by the introduction of transformers. This talk will focus on a recent iteration of transformers, Scaleformers, and the benefits it provides. While modifications of transformers have focused in recent years on improving the effectiveness of the attention mechanism or better modelling frequency information, Scaleformers focuses on making and refining multi-scale time-series forecasts. By iteratively refining a forecasted time series at multiple scales with shared weights and introducing architecture adaptations and a specially-designed normalization scheme, we are able to achieve significant performance improvements with minimal additional computational overhead. Experiments on public datasets confirm the benefits of this method, both on standard point-forecast datasets, as well as in a probabilistic forecasting setting.

Event Agenda:

• 5:30 pm - 6:00 pm - Sign in, Grab a name tag, Pizza + Network
• 6:00 pm - 6:05 pm - Welcome / Opening Remarks
• 6:05 pm - 6:35 pm - Talk 1 + Q&A
• NETWORKING BREAK
• 6:55 pm - 7:25 pm - Talk 2 + Q&A
• 7:25 pm - 8:00 pm - Wrap-up

Location:
Find Moov AI at 4115, St-Laurent Boulevard, Suite 300, Montréal, Québec, H2W 1Y8. Enter the main building doors at street level. A team member will be waiting to greet you, and directional signage will be placed throughout. *Email [melissa.stabner@borealisai.com](mailto:melissa.stabner@borealisai.com) if you require access to an elevator.