Skip to content
 

“Scalable Bayesian Inference with Hamiltonian Monte Carlo” (Michael Betancourt’s talk this Thurs at Columbia)

Scalable Bayesian Inference with Hamiltonian Monte Carlo

Despite the promise of big data, inferences are often limited not by sample size but rather by systematic effects. Only by carefully modeling these effects can we take full advantage of the data—big data must be complemented with big models and the algorithms that can fit them. One such algorithm is Hamiltonian Monte Carlo, which exploits the inherent geometry of the posterior distribution to admit full Bayesian inference that scales to the complex models of practical interest. In this talk I will discuss the theoretical foundations of Hamiltonian Monte Carlo, elucidating the geometric nature of its scalable performance and stressing the properties critical to a robust implementation.

The talk is this Thurs, 6 Apr, 1:10-2:20pm in 303 Mudd Building at Columbia.

You shouldn’t miss this one. These ideas are fundamental to Stan present and future.

4 Comments

  1. Jonathan says:

    Will you post slides? This topic interests me a lot. I’m particularly interested in the “scales to” part.

  2. Jordan Anaya says:

    A couple things:
    It appears Wansink took down his blog post, which is where he was supposed to post the reanalysis of the pizza papers (which was supposed to have been done by now).

    I’m giving a talk at UPenn on Thursday on pizzagate. I’ve already uploaded the talk to youtube, but I’m waiting on some feedback before I make the link public.

    Anything I need to make sure I do in Philadelphia?

    • Andrew says:

      Jordan:

      In your talk perhaps you can do a live demo, something along the lines of the activity on page 97 here: http://www.stat.columbia.edu/~gelman/research/published/smiley12.pdf Or maybe the point is too obvious and doesn’t need to be made. It could be enough to post that famous xkcd jelly bean cartoon along with the exchange, when the commenter wrote, “I suggest you read this xkcd comic carefully: https://xkcd.com/882/ It provides a great example of learning from a ‘deep dive’. [quoting Wansink]” and Wansink’s reply, in its entirety, was this: “Hi Anthony, I like it. Thanks for the link. (Makes me grateful I’m more of a purple jelly bean guy).” Or you could go subtle and, every time you mention Wansink, put up a photo of Diederik Stapel and see if anyone notices.

Leave a Reply