Here’s related work from the lab that we’ll be building on: bsky.app/profile/hafe...
Can we simultaneously learn transformation-invariant and transformation-equivariant representations with self-supervised learning?
TL;DR Yes! This is possible via simple predictive learning & architectural inductive biases – without extra loss terms and predictors!
🧵 (1/10)
Here’s related work from the lab that we’ll be building on: bsky.app/profile/hafe...
Colin Conwell, @emaliemcmahon.bsky.social, Akshay Vivek Jagadeesh, Kasper Vinken, @amrahs-inolas.bsky.social, @jacob-prince.bsky.social, George Alvarez, @taliakonkle.bsky.social and Marge Livingstone
Colin Conwell, @emaliemcmahon.bsky.social, Akshay Vivek Jagadeesh, Kasper Vinken, @amrahs-inolas.bsky.social, @jacob-prince.bsky.social, George Alvarez, @taliakonkle.bsky.social and Marge Livingstone