Scribd
Upload
48 views

Dynamic Programming: Reinforcement Learning Homework Assignment - Move 37

dl

Uploaded by

zentropia
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
48 views

Dynamic Programming: Reinforcement Learning Homework Assignment - Move 37

dl

Uploaded by

zentropia
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 5

Dynamic Programming

Reinforcement Learning Homework


Assignment – Move 37
Collected by Sudha Mantri (Dean, Leesburg School of AI, VA, USA)

V 1.0
Dated 9/12/2018
Table of Contents

Disclaimer...................................................................................................................................................... 3
Version History.............................................................................................................................................. 4
Value Function – Quiz ................................................................................................................................... 5
Disclaimer

I thank Siraj, all the internet sources, experts, and School of AI team members. This information is not
meant to violate any copyrights. Kindly know that I am an eternal student (always learning) and making
notes (literally). However, I wish to add that this information is meant to be used by School of AI
Deans/Wizards to help members and aspirants learn AI Free of charge. Please let me know of any
copyright violations and I shall duly oblige.

• *School of AI Deans please feel free to use this information and/or create your own
sessions/training as you deem fit

Below is the School of AI Guidelines for all the deans. Please follow them and share information if/as
need be.

https://gallery.mailchimp.com/21a24cf96b7ec98b2144fca5c/files/c9a14833-10b2-42c0-a024-
f5310cb395bc/Copy_of_School_of_AI_Meetup_Guidelines.pdf

School_of_AI_Meetup_Guidelines.pdf

3
Version History
V 1.0 9/7/2018 Initial Version (Draft) meant for School of AI Deans as reference – Use
to teach others for FREE

4
Dynamic Programming (DP) and Reinforcement Learning (RL) – Reading
Assignment

• Many reinforcement learning (RL) algorithms utilize dynamic programming methods.


Dynamic Programming methods have been researched for years to solve MDPs (Markov
Decision Processes) assuming that the agent has prefect knowledge of functions that
define an environment. In machine learning, the environment is usually formulated as
an MDP. The main difference between the classical dynamic programming methods and
RL algorithms is that RL algorithms do not assume knowledge of an exact mathematical
model of the MDP. RL algorithms target large MDPs where exact methods are not
possible. RL problems such as Q-learning fits a problem setting under Markov decision
processes because the environment is observable and an agent can iteratively and
progressively learn to improve. Agent works towards obtaining optimal discounted
reward from the environment.
I suggest the following as reading assignment. foundational material for DP in RL with
examples.
URL: https://github.com/dennybritz/reinforcement-learning/tree/master/DP/

Additional Reads:

1) Good Beginner/Midlevel Summary


a) https://towardsdatascience.com/reinforcement-learning-demystified-solving-mdps-
with-dynamic-programming-b52c8093c919
2) Comprehensive Reading:
a) http://incompleteideas.net/book/bookdraft2018jan1.pdf
b) http://www.wildml.com/2016/10/learning-reinforcement-learning/
3) Midlevel/Advanced Summary:
a) https://www.cs.cmu.edu/~katef/DeepRLControlCourse/lectures/lecture3_mdp_plannin
g.pdf
4) Advanced
a) https://djrusso.github.io/RLCourse/index

You might also like