iiinova — where your digital self keeps evolving

James Allingham

@jamesallingham.bsky.social

1.1K followers 173 following 8 posts

Research Scientist @GoogleDeepMind | Organiser @DeepIndaba | Machine Learning PhD @CambridgeMLG | 🇿🇦

Posts Replies Media

Reposted by James Allingham

Tim G. J. Rudner @timrudner.bsky.social · 05/02/2025

Submit your work to the 7th Symposium on Advances in Approximate Bayesian Inference! #AABI This year, #AABI will be co-located with #ICLR2025! Workshop Track: February 7, AoE Proceedings Track: February 7, AoE Fast Track: February 18 / March 14, AoE approximateinference.org/call/

approximateinference.org

Call for Papers

1196

James Allingham @jamesallingham.bsky.social · 05/12/2024

Thanks 🤩 Obsessing over TikZ is my guilt-free procrastination method 😂

030

James Allingham @jamesallingham.bsky.social · 05/12/2024

A big shoutout to all of my amazing collaborators who made this paper happen! @brunokm.bsky.social Shreyas Padhy, Javier Antoran, David Krueger, Richard Turner, Eric Nalisnick, and Jose Miguel Hernandez-Lobato.

030

James Allingham @jamesallingham.bsky.social · 05/12/2024

I'll keep this thread short, but if you are interested to chat further please get in touch or visit the poster at NeurIPS on Fri 13 Dec at 4:30 p.m. PST (East Exhibit Hall A-C #3710) Here are a few diagrams more diagrams from the paper to tempt you!

110

James Allingham @jamesallingham.bsky.social · 05/12/2024

Excitingly, we can also use the symmetry information learned by our SGM to improve the data efficiency of standard deep generative models (e.g., VAEs).

100

James Allingham @jamesallingham.bsky.social · 05/12/2024

Our SGM is also interpretable – we can inspect the distributions over transformations for any prototype, which tells us about our dataset, and if our SGM is learning reasonable things. E.g., 9's and 6's can be rotated into each other, and 1's can be rotated 180 deg w/o change.

110

James Allingham @jamesallingham.bsky.social · 05/12/2024

We provide experimental evidence that our SGM can learn prototypes and the distributions over transformation parameters such that the true data distribution is recovered. Here we show observations from the test set (top), prototypes (mid), and resampled observations (bot).

100

James Allingham @jamesallingham.bsky.social · 05/12/2024

We introduce our symmetry-aware generative model (SGM), in which an observation is generated by transforming an invariant latent "prototype", and a simple algorithm for learning the protos and transformation params. paper: arxiv.org/abs/2403.01946 code: github.com/cambridge-ml...

100

James Allingham @jamesallingham.bsky.social · 05/12/2024

I'll be at NeurIPS next week, presenting our work "A Generative Model of Symmetry Transformations." In it, we propose a symmetry-aware generative model that discovers which (approximate) symmetries are present in a dataset and can be leveraged to improve data efficiency. 🧵⬇️

2181