PyData Montreal Meetup #6

Join us for our 6th PyData Montreal meetup:

• "Neural Episodic Control" by Simon Ouellette
• "Beyond the basics of similarity search" by Alex Kim

----------------------------------------------------------------------------------------------------
Agenda
6:00 pm—Doors open

6:10 pm— "Neural Episodic Control" by Simon Ouellette

Abstract:
It is a well-known fact in machine learning that gradient descent-based techniques, like traditional deep learning, learn extremely slowly. It typically takes thousands or millions of examples to learn something that should only take a few samples at most.

In this talk, we will address this issue for the case of Reinforcement Learning, by discussing the topic of Episodic Control. Based on concepts of memory and state-similarity lookups, this approach seeks to reproduce what deep reinforcement learning can do, but without a gradients-based core, thus allowing it to learn concepts much faster

About Simon:
Simon Ouellette is a quantitative trader, currently spearheading the AI trading initiative at Pointus Partners. Starting his career as a computer scientist about 15 years ago, he increasingly gravitated towards data science-related roles, culminating in his current position. He also creates content about topics such as stochastic modelling, reinforcement learning and quantitative trading via his blog at http://www.simonouellette.com/

7:00 pm — Break

7:10 pm — "Beyond the basics of similarity search" by Alex Kim

Abstract:
Search and retrieval of nearest neighbours, i.e. a subset of data points similar to a query data point, is a common data science problem.
For large datasets, exhaustive search is computationally expensive.
This talk will introduce various types of approximate nearest neighbour algorithms that produce results that often "good enough" from a practical point of view while significantly improving speed.

About Alex:
Most of Alex's work experience involved solving data-sciency problems in various domains: physics, aerospace, telemetry/log analytics, image and video processing.
Lately, he has been working as an independent Data Science consultant.

8:00 pm — Break and networking