Attempts on finding interesting research questions
Exploring ideas around transfer learning and loss landscapes.
Still somewhere deep in the loss landscape…
This blog will features my second PhD Talk and the following repos:
- tflearning: A repository for finetuning, linear mode connectivity, learning dynamics and sample difficulty experiments.
- lfmd_notebooks: A collection of experiment analysis notebooks of that time.
- minml_utilities: A minimal starter framework for new ml projects at this time.
I tweeted about some results on sample difficulty:
I also stumbled across the cool "Deep learning on a data diet" paper by @mansiege and @gkdziugaite and tried to reproduce their results.
— Maximilian Beck (@maxmbeck) March 31, 2023
I could not reproduce the GraNd results either and thought this must be a bug in my code. 🪲
Very cool, that @BlackHC dug a bit deeper.⛏️👷♂️ https://t.co/UohXrzjVTK pic.twitter.com/Eyl6twBxgT