Causal Modeling in Machine Learning

150 150 The TWIML AI Podcast (formerly This Week in Machine Learning & Artificial Intelligence)

We’re collaborating with research scientist and instructor Robert Ness to bring his course sequence, Causal Modeling in Machine Learning, to the TWIML Community. Causality has become a very hot topic in the ML/AI space. In fact, it’s come up in a good number of my recent conversations, like this one with Zach Lipton. One of the challenges facing those interested in learning about causality in ML is that most resources on the topic are geared towards the needs of statisticians or economists, versus those of data scientists and machine learning engineers.

Robert, an ML research scientist at startup Gamalon and an instructor at Northeastern University, has developed a series of six course modules on Causal Modeling in Machine Learning that are designed to be more practical and accessible for data scientists and engineers. He is teaching the course live to graduate students at Northeastern University, and through our new partnership Robert will also be hosting a study group via the TWIML platform, i.e. Zoom and Slack.

The study group will provide TWIML enrollees some of the benefits of taking the course live. Robert will hold a weekly review session after each week of study in the sequence, will be available to answer questions via Slack, will personally grade submitted assignments, and will be available to assist with course homework and projects.

To learn more about the courses and study group, check out the recording from our webinar, or peruse the FAQ we’ve prepared below.

When you’re ready to enroll, you can do so at Robert’s AltDeep.ai web site. Be sure to also join the TWIML Community and the #causality_course channel on our Slack.

Frequently Asked Questions

  1. What are the courses?

    The six modules are listed on the AltDeep web site. They are:

    • Model-based Thinking in Machine Learning. Lay the foundation for causal models by deconstructing mental biases and acquiring new mental models for applied DS/ML.
    • Do Causality like a Bayesian. Continue your “mental refactoring” by developing a Bayesian mental model for machine learning.
    • How to Speak Graph; or DAG that’s a Nice Model! Become fluent in directed graphs and graph algorithms as a language of probability.
    • The Tao of Do; Modeling and Simulating Causal Interventions. Learn to build your first causal generative machine learning model using a deep learning framework
    • Applied Causal Inference; Identification and Estimation of Causal Effects from Data. Gain mastery of programmatic causal effect estimation.
    • Counterfactual Machine Learning. Implement counterfactual reasoning algorithms in automated decision-making settings in industry.
  2. Are the courses free? How much are they? How are they sold?

    These are paid courses. Robert has put a ton of work into this sequence and will be providing TWIML learners with human support as they take the courses.

    Rather than selling the modules individually, Robert sells them in two packages or “workshop options”:

    Refactored Thinking for Machine Learning and Causality, aka the “short course,” consists of the first two modules listed above. It costs $299.
    Causal Modeling in Machine Learning Track, aka the “long course” or “full sequence,” consists of all six of the modules. It costs $1,199.

    The short course is designed to lay a foundation and help you become conversationally fluent in causal ML, and the long course is designed to take you deeper into the practice of causal ML.

  3. How long will each course run? What is the level of effort expected?

    The short course will run one month, i.e. 2 weeks per module, and the long course will run four months with some of the later modules running longer.

    On time commitment, if you just want to go through lectures and videos, then the time commitment is akin to a deep read of one paper a week. If you wanted to work through code examples and assignments, then more. The course is designed to give you a level of depth that suits you.

  4. Is there a discount for TWIML participants?

    Glad you asked. Yes, to kick off this partnership, Robert has agreed to extend a 20% discount to TWIML participants who register using the links above. I suspect this is the lowest price these courses will ever be offered for.

  5. Is TWIML paid as part of this arrangement?

    Yes, we are an AltDeep / Teachable affiliate and get a commission as part of the partnership. Whatever we earn through this relationship will help support our broader community and educational efforts. That said, we would never recommend a course we didn’t think was a good use of your time and a good value.

  6. What if I start with the short course and decide I want to continue?

    Your full short course purchase will be credited to your long course purchase/upgrade. You will receive the original TWIML discount on the upgrade purchase.

  7. How long will students have access to the course materials?

    After you enroll, you will have access to the materials indefinitely.

  8. How long will the course be open? How long will the discount be available?

    While the course itself is fundamentally designed for self-paced study, with Robert running a live weekly study group, we will be ending the discount on Thursday, February 6th at 11:59 pm EST. Enrollment will be closed one week later.

  9. Will the weekly study group sessions be open to anyone?

    Robert’s weekly study group sessions are intended for enrollees and will assume that learners have at least gone over that week’s lectures at a high level.

  10. Is there a detailed syllabus?

    Yes, the syllabus will roughly follow that of Robert’s Northeastern course, which you can find here.

  11. What programming languages/frameworks are used in the course?

    The courses incorporate probabilistic programming concepts and use Pyro. From the Pyro web site:

    Pyro is a universal probabilistic programming language (PPL) written in Python and supported by PyTorch on the backend. Pyro enables flexible and expressive deep probabilistic modeling, unifying the best of modern deep learning and Bayesian modeling.

  12. Where can I learn more about Robert and the course?

    You can check out a replay of our recent webinar with Robert above. (Please be sure to like and subscribe while you’re there!) You can also check out my recent podcast, Causality 101, with Robert, here.

  13. What can I expect from the weekly online sessions?

    The weekly online sessions are live study group sessions presented by Robert, the instructor and author of the Causal Models in Machine Learning courses. At the sessions, Robert will present a summary of that week’s lecture and open the floor for student Q&A.

  14. What if I cannot participate in the weekly online sessions?

    The weekly study group sessions will be recorded and will be available to TWIML enrollees.

  15. What is the refund policy?

    There is a 30-day refund policy on the courses.

5 comments

Leave a Reply

Your email address will not be published.