Help reproducing Figure 8 from the paper

Nobody has claimed this yet.

Hi,

This is super interesting work and thank you for sharing the code.

I'm trying to build some intuition on how the anchors should be chosen and in the process I am trying to reproduce Fig. 8 in your [paper (arXiv:2207.07235)](https://arxiv.org/abs/2207.07235). However, I'm having some trouble and was hoping you could help. Specifically, I'm unable to train a model that predicts the large uncertainty between training samples when the number of samples is low i.e 5 samples.

I adapted the [demo notebook](https://github.com/LLNL/DeltaUQ/blob/main/demo/demo.ipynb) and can share that but I don't suppose you have the code to generate those plots at hand and share them perhaps?

Many thanks again!

Start with demo/demo.ipynb and compare its training setup with the procedure described for Figure 8 in the paper. Reproduce the five-sample case and identify why the predicted uncertainty differs; done means the Figure 8 plots can be generated or the missing reproduction steps are documented.

Written by the indexing model from the issue text.