whest_phase1_writeup_skibidi_toilet.pdf (482.1 KB)
6 Likes
With the hot/cold approach managed to reduce multiplier by 6% (with similar final MSE, this on a held out fresh set of 100 nets). Nice idea!!
2 Likes
Unreal! Excellent work and innovative too. At the end I think you and someone else were right ahead of me on the leaderboard, every single of the ai tools I was using said this is the best you’re going to get, not possible to move higher, no /brainstorming, no articles, your crazy ideas or books will work other than maybe a potential dynamic estimator, I was like this dude clearly is not oversampling, he’s just got something figured out–they came to the conclusion you weren’t real except claude, claude just told me to go to sleep and didn’t address 
1 Like