Skip to content

Latest commit

 

History

History
 
 

week5

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
 
 
 
 

Materials

More materials

Homework

As usual, "lasagne way" and "other way"

Lasagne way

Basically go to the notebook and follow what's inside.

Other way

This week's task is to implement (and hopefully compare) target networks, double DQN and/or duelling DQN and training on atari breakout.

Implementing prioritized experience replay or bootstrap dqn or any other cool stuff yields you bonus points. You can also choose a different environment if you have issues with breakout, but don't get too complicated. E.g. your DQN will likely fail on Montezuma Revenge unless you do weird stuff with reward function.

We recommend you to upload your results to OpenAI gym and fit your solution in a notebook (ipython/torch/r) unless your framework is incompatible with that. In the latter case, please supply us some notes on what code lies where.

Again,we recommend you to read the lasagne/agentnet assignments briefly to get the grasp of what parameters to start from.

Bonus assignments remain exactly the same as in the first track.

Blindly copy-pasting code from any publically available demos will result in us interrogating you about every signifficant line of code to make sure you at least understand (and regret) what you copypasted.