viddesh1 / rl Goto Github PK

After the agent learn the q-table policy should be dump in to pickle object file. So, that it is easy to un-pickle and render visualization of frozen lake environment.

• Make environment of Frozen-lake_v1.
• Read the pickle file.
• Choose the maximum action form the Q-table
• Start the episodes and render the UI.
• Add necessary docstring and comments for code readibility.

Hi @shruti-madhav,

Could you like to work on this issue task after .ipynb or .py file is successfully created in remote repository?

OR

Should I take care?

Kindly let me know so that I can assign the task to you.

Adding @Ankita01K for visibility.

Thanks!
Viddesh

There was some issue when importing the package of gym. May be the issue was because using of come class or function keyword of the gym packages.

@Ankita01K @shruti-madhav Have a run on your local machine to check if everything in working correctly in your laptops.

Create a function nd call action function and Q-table update function until the agent learn a q-table policy to reach from the start node to the goal node.

A video representation of the working mini project will look good in the markdown file.

Since, the learning of the machine learning model takes a significant time to compute. I will put the output in comments.

More consise information can be added for clarification.

Adding the requirements.txt file will be very help for installation of the necessary packages and any major dependencies for the project.

Requirement file should contain below packages one major dependencies of pygame:-

• requirements.txt

numpy==1.23.5
gym==0.24.0
pygame==2.3.0

Create a function which take the current state that agent is in. Based on the state an episilon-greedy policy to be used for taking random uniform action in the environment else action should be taken by agent by following a greedy policy of Q-table.

The Problem is when we run the first cell and last cell of the Mini_Project.ipynb file. We are unable to close the pygame window.
As a result of which we have to forcefully kill the kernel and run again first and last cell.

How to resolve this issue?

• Import pygame in the first cell like --> import pygame
• Print the pygame version --> print("Pygame Version: ", pygame.version)
• We would need to use pygame.quit() at the end of the line outside of the forloop in the last cell of Mini_Project.ipynb. For closing the pygame window as gym uses pygame as internal dependencies.

When Sending for Pull Request kindly tag this issue and checkmark all the checkbox. Thanks!

Why not to have a file structure representing the file names and there information in the form of tree structure. It will provide more insights on how the file structure is created and why it is in that way. Also, Meta information of file may be helpful.

Interactive python notebook (.ipynb) file need to be created so that we can test the code. As it provides more flexibility than that of simple python (.py) file.

If there is any need to create a python file we can create it and code can we maintain in the same file.

### Description of 3 sub task :-

• Creating an instance of the discrete frozen lake environment.
• Setting the necessary hyper parameters for reinforcement learning.
• Initializing Q-table with all zeros.

Previously the packages table for the project was like below:-

Updated table would look like below:-

This function should take as parameters (current_state, future_state, reward, action).
This should should use Q learning formula.

I think that a tabular representation of the technology that we used to the develpment of the mini-project in the markdown file of README.MD will look good.

Create a "Go to Top Button" at the very end of the Readme.md file. So, that when the user read through the document in readme file from top to buttom. He or She does not have to scroll back to top again to read again etc.

Adding comments to the packages so that. It can be easy to understand

Printing the version of the packages used will help in understanding which packages are used in the development of the code.