Generalization – Interpolation – Extrapolation in Machine Learning: Which is it now!?

A summary of the Generalization – Interpolation – Extrapolation debate in machine Learning. “Learning in High Dimension Always Amounts to Extrapolation” explained.

Check out Y. LeCun and R. Balestriero defend their opinion ‪@MachineLearningStreetTalk‬ : 👉   • #61: Prof. YANN LECUN: Interpolation,...   💪

Paper:
📜 Balestriero, Randall, Jerome Pesenti, and Yann LeCun. "Learning in High Dimension Always Amounts to Extrapolation." arXiv preprint arXiv:2110.09485 (2021). https://arxiv.org/abs/2110.09485
🔗 Twitter announcement by Y. LeCun:   / 1450560732483948545  
🔗 Response of F. Chollet:   / 1450524400227287040  
Older Twitter fights: 🍿
🔗 Y. LeCun claiming that in high dimensions everything is extrapolation.   / 1409940043951742981  
🔗 Gary Markus claiming the opposite:   / 1411401507610796032  
🔗 “You Don’t Understand Neural Networks Until You Understand the Universal Approximation Theorem” Blog Post by Andre Ye:   / you-dont-understand-neural-networks-until-...  
🔗 Reddit discussion about the paper:   / r_learning_in_high_dimension_always_amount...  

Thanks to our Patrons who support us in Tier 2, 3, 4: 🙏
donor, Dres. Trost GbR, Yannik Schneider
➡️ AI Coffee Break Merch! 🛍️ https://aicoffeebreak.creator-spring....

Outline:
00:00 Paper warning
01:44 It’s interpolation
03:05 No, it’s extrapolation
06:38 Now it’s interpolation again
08:08 What is it??

▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀
🔥 Optionally, pay us a coffee to help with our Coffee Bean production! ☕
Patreon:   / aicoffeebreak  
Ko-fi: https://ko-fi.com/aicoffeebreak
▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀

🔗 Links:
AICoffeeBreakQuiz:    / aicoffeebreak  
Twitter:   / aicoffeebreak  
Reddit:   / aicoffeebreak  
YouTube:    / aicoffeebreak  

#AICoffeeBreak #MsCoffeeBean #MachineLearning #AI #research​

Music 🎵 : Cha Da Cup Riddim - Konrad OldMoney
Video contains the rock emojis designed by OpenMoji – the open-source emoji and icon project. License: CC BY-SA 4.0