Melbourne, Australia. Feb 7, 2018. Understanding lp__: proportionality constants and (automatic) transforms


#1

There’s a Stan meetup in Melbourne and I’m speaking at it! It’s on Wednesday, February 7, 2018.

Title:

Understanding lp__: proportionality constants and (automatic) transforms

Description:

This is going to be an informal and hands-on talk. Please be prepared to write some Stan, R, or at least do some arithmetic on paper.

There’s a lot going on in Stan. How lp__ is computed is part of what makes Stan efficient. We’ll discuss the practical aspects of only computing things up to a proportionality constant, the importance of the automatic transforms, and how to apply Jacobians for transforms.

Register here:

Hope to see some of you there!


#2

Neat. I love how you’re going through all the FAQ topics we get endless questions about.

This gets especially tricky when there’s truncation or other normalizing constants which can sometimes be dropped and sometimes not. Similarly for Jacobians for linear transforms (constant) vs. non-linear (not constant).


#3

I’ll improve the talk and give it again back home sometime soon.

Yup. It’s all very clever.


#4

Having worked with this for so long, it seems natural to me. I forget how strange it all is for new people because it’s very convoluted compared to most programming languages.