only compute Q-values for feasible actions (just because DQN computes Q-values for a hardcoded set of actions doesn't mean you have to); 3.
use Achiam's "Constrained Policy Optimization" https://arxiv.org/abs/1705.10528 ; 4.
The app was created by three University of Iceland software engineering students for a contest calling for "new creative uses" of the Islendingabok, or Book of Icelanders, an online database of residents and their family trees stretching back 1,200 years.
Arnar Freyr Adalsteinsson, one of the trio, said it allows any two Icelanders to see how closely related they are, simply by touching phones."A small but much talked about feature is the loosely translated 'Incest Prevention Alarm' that users can enable through the options menu which notifies the user if the person he's bumping with is too closely related," Adalsteinsson said. Their descendants built a small, relatively homogenous and — crucially — well-organized country, home to the world's oldest parliament and devoted to thorough record-keeping."The Icelandic sagas, written about 1,000 years ago, all begin with page after page of genealogy.
The problem is that there's so little inductive bias there that it takes forever to get anywhere useful, you have to evolve far too many samples.
With the evolving CNNs, researchers are already joking about how you would need a nuclear reactor to reproduce some of these papers which train thousands of CNNs, so the inductive bias is really important in keeping samples down to a feasible level.
Currently available for Android phones, it has been downloaded almost 4,000 times since it was launched earlier this month.Some are hailing it as a welcome solution to a very Icelandic form of social embarrassment."Everyone has heard the story of going to a family event and running into a girl you hooked up with some time ago," said Einar Magnusson, a graphic designer in Iceland's capital, Reykjavik."It's not a good feeling when you realize that girl is a second cousin.