HRL
HRL copied to clipboard
NotAnyMike
Hierarchical Reinforcement Learning (batteries included)
that could allow the agent to go through x intersections and count the ones taken correctly, for this to work the discount has to be very very high (i.e. 1...
This will make Take_center much more useful. **There are policies Turn_left/right already trained on only left and right and not straight
so we can only specify the number of the experiment, because it already ignores that in the version of the weights
Because then, in reality running for 3m steps is 3x4env steps
add sections: - [ ] Intro: show all the cool stuffs if has in a video, few images and few text - [ ] Custom env: Show how easy is...
Little information, we know that it happens when we repeat several times, the track creation, x10, but it has failed around 20-30% of the whole training steps (3m) --- update:...
Metadata
Owner
Metadata
Hierarchical Reinforcement Learning (batteries included)