MACHINE LEARNING

The diffusion model is built on top of our neural operator. You cannot compare them directly since neural operators are deterministic. A natural comparison is with @CristopherSalvi SDE formulation but that assumes a specific form of stochasticity in equations

→ View original post on X — @animaanandkumar

T-STAR is a style transfer approach that uses AMR graphs for intermediate representations. The first of it's kind, T-STAR yields high content preservation with negligible accuracy loss. Drop by the @emnlpmeeting Google booth at 3:30pm today to hear @JangraAnubhav talk about it!

→ View original post on X — @googleai

#NeurIPS2022, the longest machine learning conference ever pic.twitter.com/ku7PVdclw7

— Gautam Kamath (@thegautamkamath) 10 décembre 2022

#NeurIPS2022, the longest machine learning conference ever

→ View original post on X — @thegautamkamath

Anyone who claims that the successes of modern machine learning are just due to scaling up needs to explain away each of these essential pieces:
– Adam
– Batch norm
– Resnets
– Contextual embeddings
– Transformers

→ View original post on X — @pmddomingos

It's remarkable both that backprop is just a simple optimization scheme and that it took a psychologist to come up with it.

→ View original post on X — @pmddomingos