4 ms·
I got my PhD in this field and now develop methods like these. They can work, but the field is still immature - there are plenty of models in circulation that h
by comicjk 9y ago
I got my PhD in this field and now develop methods like these. They can work, but the field is still immature - there are plenty of models in circulation that haven't been proved to deliver any useful information. Although I agree with the list of improvements suggested by the authors, I would put "more validation" above all of them.
- stochastic_monk 9y agoI think that high volume validation datasets for predicted/measured behaviors will be important. Maybe automation in experiment is necessary to get to that size, but if you can't verify it, it's not science.
- nether 9y agoIs there any way a non-PhD machine learning engineer who took a few materials classes in college can contribute to this work? It seems profoundly important.