Jack Medley
@jackmedley.bsky.social
I've just started playing around with numpyro's SVI using the bike's dataset I came across in @juanitorduz.bsky.social's nice blog post. It's over-dispersed count data so I'm using NB2... Does anyone have any tips on how to stop SVI cheating by sending the concentration parameter to zero??
October 4, 2025 at 9:31 AM
I've just started playing around with numpyro's SVI using the bike's dataset I came across in @juanitorduz.bsky.social's nice blog post. It's over-dispersed count data so I'm using NB2... Does anyone have any tips on how to stop SVI cheating by sending the concentration parameter to zero??
For my first bluesky post I'd like to share my four part series looking at recursive Bayesian learning.
Starting with a simple linear regression and ending at non-linear models in non-stataionary environments!
Post one is here: www.jackmedley.com/posts/2025-0...
Starting with a simple linear regression and ending at non-linear models in non-stataionary environments!
Post one is here: www.jackmedley.com/posts/2025-0...
Online Bayesian Learning Part 1: Exact recursive inference
Part 1 of a series of posts looking into the problem of ‘online’ learning using Bayesian methods
www.jackmedley.com
September 26, 2025 at 5:25 AM
For my first bluesky post I'd like to share my four part series looking at recursive Bayesian learning.
Starting with a simple linear regression and ending at non-linear models in non-stataionary environments!
Post one is here: www.jackmedley.com/posts/2025-0...
Starting with a simple linear regression and ending at non-linear models in non-stataionary environments!
Post one is here: www.jackmedley.com/posts/2025-0...