The tldr is in the abstract: "In deep learning terminology, this amounts to saying that besides a negligible set, all functions that can be implemented by a deep network of polynomial size, require an exponential size if one wishes to implement (or approximate) them with a shallow network."
That's a very interesting result (assuming it's correct).
It certainly agrees with intuition based on analogy to boolean circuits where, for example, the parity function requires exponential circuit size for shallow circuits but only linear size for deep circuits, but I haven't heard of a proof of this for NN's before.
It seems that the circle line is missing. I guess that all its stations are shared with other lines, so the network is not missing much, but still it would be nice for the sake of completeness...
And agree with the previous comment, it would be awesome if the map was truly "live".
Switzerland definitely ranks very high in all your factors.
It's one of the safest countries for investors, has a growing startup scene, has two of the top engineering schools in Europe/World (ETHZ, Zurich and EPFL, Lausanne), has the highest rate of Nobel prizes per capita, very solid educational and healthcare systems, amazing quality of life (Zurich, Geneva, etc, often rank as the top cities in the world), and last but not least: you have amazing lakes and mountains everywhere!
Agree, Switzerland seems definitely one of the top countries in the world which keeps itself intentionally under the radar. As Taleb put it, in addition to all the obvious benefits, Switzerland seems to be an antifragile country.
In any case, as far as I know Switzerland makes it pretty hard for people to immigrate there. Do you happen to live there? What would you say would be the most effective way to immigrate there?
It's easy for members of the European Union as there are/were bi-lateral agreements. For other nationalities I think it's harder, as there are some quotas. You basically need to find a job first, but it's doable.
Yes! So now you can guess why he was loading a specific different parameter file for each picture!
Translation: someone tried many different parameters for each image in this demo and them manually selected the ones that make a better result.
This might still be useful if there is a good UI for users to do the same.
