Thinking about causality in ML: Bayesian Networks

In this QuantumBlack Asia-Pacific virtual meet up, we will discuss and show how to practically use Bayesian Networks, and why this is the preferred machine learning method.

The virtual session will begin with a discussion on the importance of causal models in machine learning, some practical demonstrations leveraging Bayesian Networks, and the open-source software package, CausalNex. This will be followed by a presentation on Bayesian Networks from industry expert, Ross Pearson, ICT Director, Monash University.

Talk 1 - Daisy Wood (Melbourne), Gabriel Azevedo Ferreira (Singapore), Richard Oentaryo (Singapore)

Consider an organisation, government or person wants to make a decision, using historical data. In this type of analysis the commonly known fact that “correlation does not imply causation” comes to life. It is crucial to distinguish between events that cause existing inefficiencies and those that merely correlate. Causal inference aims to determine which available controls drive specific outcomes. Many machine learning approaches disregard causal inference, despite a wide range of approaches to causal inference have been proposed in the literature. This talk will discuss the importance of causal models, as well as some practical approaches leveraging Bayesian Networks, and open source software package, CausalNex.

Talk 2 - Ross Pearson (Monash University)

What are Bayesian Networks? How do you use them practically? and when are they preferable over other contemporary ML methods? Purely data-driven techniques typically lack a significant element needed to solve complex solutions and that is the causal story. The causal story (plus data) opens up a line of questioning that cannot easily be answered with data alone (for example counterfactual testing). Bayesian Networks are ML-driven methodology that systematically incorporates the causal story for a problem domain. Ross will be presenting Bayesian Networks from a more applied (and less theoretical) perspective.

This talk is most relevant for data scientists, translators, and analytics managers.

Due to the COVID-19 virus, we will be running this event virtually. We are running one meeting regionally and all are welcome. We are looking forward to meeting everyone via Zoom and hope this will allow more people to enjoy the discussion.

Zoom link and password: To be shared closer to the date of the event, check back for details.

--------------------

About our speakers:

Daisy Wood is a Data Scientist at QuantumBlack. She has a master’s in data science and specialises in Statistics and Probability theory. Her role in QB focuses on developing optimisation and predictive models in a Bayesian framework.

Ross Pearson is an ICT Director and tertiary research manager that specialises in the delivery of digital transformation, IT management and artificial intelligence research. This includes managing or advising on large scale AI-based research projects such as IARPA CREATE, CDAP COVID Intelligence, Turning Point (Google) and DARPA SCORE. He is also a PhD candidate with an area of focus in Bayesian Network Technology, Data Science and Artificial Intelligence (AI).

--------------------

By participating in the meetup you agree that McKinsey & Company may (i) videotape, audiotape, photograph, or otherwise record your name, voice, or image, and (ii) use and distribute such videotapes, audiotapes, photographs or recordings of your name, voice or image in any texts, videos, and other materials that McKinsey may make available through websites and social media to its employees or any third parties.
If you do not wish to be recorded please inform the event organiser in advance.