You can add location information to your Tweets, such as your city or precise location, from the web and via third-party applications. You always have the option to delete your Tweet location history. Learn more
I was more thinking that it is a problem you need to be wary of. The main objection to extrapolation is that the output may not be meaningful, so the model having some predictive uncertainty quantification may be useful.
Indeed, it is typically not meaningful -- unless you have provided the model will enough information to interpret novel areas of experience space given the data it has available. It virtually never happens with real-world deep learning.
is there any good work on detecting when you have stopped interpolating the data on the manifold and started extrapolating away from it (and found a novel area)?
How about Bayesian neural networks?
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.
