Adding CIFAR-10 DQN experiment

[mordred-skywalker]

Reference issue

Contributes experiment related to #289

Type of change

Documentation - added experiment

What does this implement/fix?

This PR adds a Deep Q-learning experiment notebook on the task of image classification using CIFAR-10 dataset.

Additional information

[codecov]

Codecov Report

Merging #525 (e35eeb3) into staging (31b0f3d) will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff            @@
##           staging     #525   +/-   ##
========================================
  Coverage    90.09%   90.09%           
========================================
  Files            7        7           
  Lines          404      404           
========================================
  Hits           364      364           
  Misses          40       40           

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 31b0f3d...e35eeb3. Read the comment docs.

[PSSF23]

You are mixing keras and tensorflow.keras here. The later one should be always used. Also, please rename notebook to cifar_q_learning.

[mordred-skywalker]

You are mixing keras and tensorflow.keras here. The later one should be always used. Also, please rename notebook to cifar_q_learning.

Fixed!

[PSSF23]

Not completely I think. You don't need to install keras from GitHub anymore right?

[mordred-skywalker]

Not completely I think. You don't need to install keras from GitHub anymore right?

I don't think I have installed keras package in the notebook. Are you referring to the keras-rl2 package? That is an independent package for deep reinforcement learning algorithms. I am attaching the link here. (https://github.com/taylormcnally/keras-rl2)

You can also refer to Sarah's PR since we are using the same imported packages.

https://github.com/neurodata/ProgLearn/blob/staging/docs/experiments/Deep%20Q-learning%20in%20XOR-XNOR.ipynb

[PSSF23]

@mordred-skywalker cool! Thanks for pointing that out. They use tensorflow.keras as well so it's all good. I will take another look over the notebook.

[mordred-skywalker]

@mordred-skywalker cool! Thanks for pointing that out. They use tensorflow.keras as well so it's all good. I will take another look over the notebook.

[mordred-skywalker]

@mordred-skywalker cool! Thanks for pointing that out. They use tensorflow.keras as well so it's all good. I will take another look over the notebook.

Sure! please let me know if other changes are required.

[PSSF23]

A lot of imported packages are not used. You can remove them and make the import code block shorter. Other than that LGTM.

[PSSF23]

Hmmm, it seems that you also need to work on GitHub skills as well, so that you don't delete & re-upload files each time, resulting in unnecessary commits. Try watching some video tutorials on how to commit & push your changes with GitHub Desktop or command line.

[mordred-skywalker]

Hmmm, it seems that you also need to work on GitHub skills as well, so that you don't delete & re-upload files each time, resulting in unnecessary commits. Try watching some video tutorials on how to commit & push your changes with GitHub Desktop or command line.

Yes, I am relatively new to Github but I'll look into that :)

[PSSF23]

There are still unnecessary imports like pandas, random, & shuffle. Look through each line and just search in your file. If you did not use it, discard it.

[PSSF23]

Hmmm, it seems that you also need to work on GitHub skills as well, so that you don't delete & re-upload files each time, resulting in unnecessary commits. Try watching some video tutorials on how to commit & push your changes with GitHub Desktop or command line.

Yes, I am relatively new to Github but I'll look into that :)

You have been using it for a semester. Better make improvements before the next~

[mordred-skywalker]

There are still unnecessary imports like pandas, random, & shuffle. Look through each line and just search in your file. If you did not use it, discard it.

Deleted them in the new commit.