3 ms·
I am surprised that the author left out another mitigation. To build solutions (models) that are constructed to be more transferable (amenable to out of domain
by slashdave 2y ago
I am surprised that the author left out another mitigation. To build solutions (models) that are constructed to be more transferable (amenable to out of domain data). For example, in machine learning, using physics informed models. Perhaps this is simply a sign that the author is a proponent of generic, deep-learning.
- yldedly 2y agoMost people in ML, even if they are very proficient, don't understand why models should generalize out of domain. They just don't think about it.