Name: 20. Can we extrapolate to extraordinary materials? | Videos | Gan Jing World - Technology for Humanity | Video & Movie Streaming
Uploaded: 2023-10-07T14:29:07.195Z
Duration: 16 min 15 s
Description: A common criticism of machine learning is extrapolation beyond the training set. Is this really poss... | Videos | Gan Jing World - Technology for Humanity | Video & Movie Streaming

Taylor Sparks

https://video4-us-east.cloudokyo.cloud/video/v6/4f/3f/ac/4f3faca5-a8d7-4df5-bc01-cd7d8a178d87/master.m3u8

A common criticism of machine learning is extrapolation beyond the training set. Is this really possible, or not? In this video, I explain how we tested to prove extrapolation to extraordinary materials *is* possible! Furthermore, we find 1. simple linear models outperform more complicated models, 2. classification outperforms regression, and 3. gaps in data (target value, missing elements, missing structures) don't prevent extrapolation.

Check out the whole materials informatics series at https://youtube.com/playlist?list=PLL0SWcFqypCl4lrzk1dMWwTUrzQZFt7y0 with workbooks and course notes available at https://github.com/sp8rks/MaterialsInformatics