A highlight of the #NeurIPS2023 meetup in Cambridge for me was Carl Rasmussen's keynote on "Halting Climate Change". Spoiler: it has nothing to do with AI. Worth a watch: youtube.com/watch?v=naFaQsFx…

Introducing a novel Langevin-based conditional sampler, our model scales to fit complex scalar/vector fields in both Euclidean & manifold codomains, exemplified by forecasting tropical hurricanes as per the illustrated figure.

We achieve this by first extending diffusion models to infinite-dimensional function spaces by a) constructing a noising process which admits, as limiting distribution, a Gaussian process that is group invariant, and b) approximating the score with an equivariant neural network.

As shown in the figure, these feature fields transform in a particular way when acted on by group elements such as translation or rotations. Often the frame used to represent these fields is arbitrary, and as such one ought to encode this invariance into the model

Our work presents a general-purpose probabilistic model over feature fields, which are functions often encountered in physical sciences, such as temperature maps, wind fields, stress tensors, etc.

as well as a keynote presentation by prof Carl Rasmussen on his latest research on climate change policy. Finally, prof @lawrennd will chair a panel discussion on "AI for Social Good".

Apart from posters, the day will feature two spotlight talk sessions...

Happy to announce the 3rd NeurIPS @ Cambridge event on Friday, Dec 8th! Join us for an in-person gathering to discuss the latest in machine learning research. Find all details and register at neuripsmeetupcambridge.info. #NeurIPS2023 #MachineLearning #Cambridge

Lol. I also have to thank T-Rex for my first job. Learned to play the game (from raw pixels in a screen capture) using deep Q-learning: github.com/vdutor/TF-rex

Kick-Off @CambridgeEllis machine learning summer school 🎒