Completed
06:28 How to think about uncertainty while doing machine learning
Class Central Classrooms beta
YouTube videos curated by Class Central.
Classroom Contents
Uncertainty in Materials Science Property Prediction - The Good, The Bad, and The Uncalibrated
Automatically move to the next video in the Classroom when playback concludes
- 1 00:00 Uncertainty in Materials Science Property Prediction
- 2 01:17 Overview
- 3 01:47 Uncertainty 101
- 4 01:50 Machine Learning for Materials Property Prediction
- 5 02:45 Do your model parameters have uncertainty? Do your data points?
- 6 03:04 If your model parameters do not have uncertainty…
- 7 03:50 If your model paramters do have uncertainty..
- 8 04:29 Frequentist & Bayesian Statistics https://xkcd.com/1132
- 9 05:55 If neither of these feel like a good fit….
- 10 06:05 CONGRATULATIONS!!!
- 11 06:28 How to think about uncertainty while doing machine learning
- 12 07:47 Different Kinds of Uncertainty…
- 13 09:39 …Lead to Different Measurements
- 14 13:27 Uncertainty Characterization Workflow
- 15 13:32 Uncertainty Characterization
- 16 16:08 State of the Art for Quantifying Uncertainty
- 17 18:17 Total Variance for a Single Sample
- 18 19:04 Uncertainty Decomposition Through Variance Conservation Assumption: variance and uncertainty are proportional
- 19 19:31 Epistemic Uncertainty
- 20 20:06 Aleatoric Uncertainty is the Bad Kind
- 21 20:36 Revisiting Model Calibration
- 22 21:24 Model Calibration for Regression Models
- 23 22:13 Case Study: Leave-one- element-out for Epistemic Uncertainty
- 24 22:16 Setting Up the Experiment
- 25 23:13 Model Generalizes Differently Depending on Omitted Element
- 26 23:58 Random Forest Prediction Results
- 27 25:41 Random Forest Uncertainty Results
- 28 27:21 Deeper Dive: Accelerated Materials
- 29 28:33 Coding Exercise
- 30 47:09 Thank you so much!
