Assignment 5

CSIE 5431 - Applied Deep Learning, instructed by Yun-Nung (Vivian) Chen

Assignment Release


  • implement Deep Reinforcement Learning algorithm
  • Acheiving weak baseline will get 4%
  • Achieving strong baseline will get 4% more
  • 6% will be graded via report
  • 1% will be graded from your format


12/29/2016 9:00am

  • Competition end
  • Start to count free day quota

Assignment 5 (15%)

Free HTML5 Template by

Assignment 5 has only one task: Atari game playing. Details are shown below.

Game Playing (8%)

Assignment 5 is to implement Deep Reinforcement Learning.
You are supposed to implement Deep Q Learning to play Atari game: Lbreakout

Github :

TAs will import and your Github should contain following files:
' ADL2016/hw5/ '
' ADL2016/hw5/best_model.ckpt '
' ADL2016/hw5/report.pdf '
' ADL2016/hw5/ '
Make sure your code contain those files mentioned above.

Dataset :

For the function you may use : please clone the code from here
For more information and details, please see homework 5 slides here

Evaluation :

TA will evaluate your model by import and test it with several episode.

Assignment 5 Grading Policy :

- achieving weak baseline gets 4%
- achieving strong baseline get 4% more

Report (6%)

Your report 'report.pdf' should follow some rules :
- Answer 3 questions we provided.(see HW slides)
- Up to 2 pages.(include figures)
- Fontsize : 12pt.

Format (1%)

- If any wrong format or error occurred, that part will get 0%


- There's "no" need to submit your code to codalab, just upload your files to Github


- For more details, please see homework 5 slides here


Any frequent question will be shown here. Post your question on FB Group or contact with TA via e-mail:

Assignment Submission


Open-source your code.


No need to submit your answer to CodaLab

homework 5 slides

Deadline 設定為12/29上課前,在此之前Assignment皆能夠按照規則繳交。

Teaching Assistants

Free HTML5 Template by


General questions; GitHub submission.

Free HTML5 Template by


General questions; CodaLab submission.