Reinforcement Learning & Deep RL Python(Theory & Projects)

Reinforcement Learning: Deep Q-Learning, SARSA, Deep RL, with Car Racing and Trading Project and Project and Interview

What you'll learn
● The introduction and importance of Reinforcement & Deep Reinforcement Learning
● Practical explanation and live coding with Python
● Deep Reinforcement Learning applications
● Q-Learning using Python
● SARSA using Python
● Random Solutions using Python
● Hyper-parameters of Deep RL
● MDP
● Mini Project (Frozen Lake) using Python
● Open AI GYM
● Intro to Deep Learning
● Deep Learning Fundamentals
● Mini Project (CIFAR) using Pytorch
● Fundamentals of DQN
● Cart-Pole from Scratch Project using Python
● Stable Baseline 3
● Cart-Pole from Scratch Project using Stable Baseline 3
● Car Racing Game Project using Stable Baseline 3
● Trading Bot Project using Stable Baseline 3
● Interview Preparations

Requirements
● Prior knowledge of Python.
● An elementary understanding of programming.
● A willingness to learn and practice.
Description
Comprehensive Course Description

Reinforcement Learning (RL) is a subset of machine learning. In the RL training method, desired actions are rewarded, and undesired actions are punished. In general, an RL agent can understand and interpret its environment, take actions, and also learn through trial and error.

Deep Reinforcement Learning (Deep RL) is also a subfield of machine learning. In Deep RL, intelligent machines and software are trained to learn from their actions in the same way that humans learn from experience. That is, Deep RL blends RL techniques with Deep Learning (DL) strategies.

Deep RL has the capability to solve complex problems that were unmanageable by machines in the past. Therefore, the potential applications of Deep RL in various sectors such as robotics, medicine, finance, gaming, smart grids, and more are enormous.

The phenomenal ability of Artificial Neural Networks (ANNs) to process unstructured information fast and learn like a human brain is starting to be exploited only now. We are only in the initial stages of seeing the full impact of the technology that combines the power of RL and ANNs. This latest technology has the potential to revolutionize every sphere of commerce and science.

How Is This Course Different?

In this detailed Learning by Doing course, each new theoretical explanation is followed by practical implementation. This course offers you the right balance between theory and practice. Six projects have been included in the course curriculum to simplify your learning. The focus is to teach RL and Deep RL to a beginner. Hence, we have tried our best to simplify things.

The course ‘A Complete Guide to Reinforcement & Deep Reinforcement Learning’ reflects the most in-demand workplace skills. The explanations of all the theoretical concepts are clear and concise. The instructors lay special emphasis on complex theoretical concepts, making it easier for you to understand them. The pace of the video presentation is neither fast nor slow. It’s perfect for learning. You will understand all the essential RL and Deep RL concepts and methodologies. The course is

• Simple and easy to learn.

• Self-explanatory.

• Highly detailed.

• Practical with live coding.

• Up-to-date covering the latest knowledge of this field.

As this course is an exhaustive compilation of all the fundamental concepts, you will be motivated to learn RL and Deep RL. Your learning progress will be quick. You are certain to experience much more than what you learn. At the end of each new concept, a revision task such as Homework/activity/quiz is assigned. The solutions for these tasks are also provided. This is to assess and promote your learning. The whole process is closely linked to the concepts and methods you have already learned. A majority of these activities are coding-based, as the goal is to prepare you for real-world implementations.

In addition to high-quality video content, you will also get access to easy-to-understand course material, assessment questions, in-depth subtopic notes, and informative handouts in this course. You are welcome to contact our friendly team in case of any queries related to the course, and we assure you of a prompt response.

The course tutorials are subdivided into 145+ short HD videos. In every video, you’ll learn something new and fascinating. In addition, you’ll learn the key concepts and methodologies of RL and Deep RL, along with several practical implementations. The total runtime of the course videos is 14+ hours.

Why Should You Learn RL & Deep RL?

RL and Deep RL are the hottest research topics in the Artificial Intelligence universe.

Reinforcement learning (RL) is a subset of machine learning concerned with the actions that intelligent agents need to take in an environment in order to maximize the reward. RL is one of three essential machine learning paradigms, besides supervised learning and unsupervised learning.

Let’s look at the next hot research topic.

Deep Reinforcement Learning (Deep RL) is a subset of machine learning that blends Reinforcement Learning (RL) and Deep Learning (DL). Deep RL integrates deep learning into the solution, permitting agents to make decisions from unstructured input data without human intervention. Deep RL algorithms can take in large inputs (e.g., every pixel rendered to the user’s screen in a video game) and determine the best actions to perform to optimize an objective (e.g., attain the maximum game score).

Deep RL has been used for an assortment of applications, including but not limited to video games, oil & gas, natural language processing, computer vision, retail, education, transportation, and healthcare.

Course Content

The comprehensive course consists of the following topics

1. Introduction

a. Motivation

i. What is Reinforcement Learning?

ii. How is it different from other Machine Learning Frameworks?

iii. History of Reinforcement Learning

iv. Why Reinforcement Learning?

v. Real-world examples

vi. Scope of Reinforcement Learning

vii. Limitations of Reinforcement Learning

viii. Exercises and Thoughts

b. Terminologies of RL with Case Studies and Real-World Examples

i. Agent

ii. Environment

iii. Action

iv. State

v. Transition

vi. Reward

vii. Quiz/Solution

viii. Policy

ix. Planning

x. Exercises and Thoughts

2. Hands-on to Basic Concepts

a. Naïve/Random Solution

i. Intro to game

ii. Rules of the game

iii. Setups

iv. Implementation using Python

b. RL-based Solution

i. Intro to Q Table

ii. Dry Run of states

iii. How RL works

iv. Implementing RL-based solution using Python

v. Comparison of solutions

vi. Conclusion

3. Different types of RL Solutions

a. Hyper Parameters and Concepts

I. Intro to Epsilon

II. How to update epsilon

III. Quiz/Solution

IV. Gamma, Discount Factor

V. Quiz/Solution

VI. Alpha, Learning Rate

VII. Quiz/Solution

VIII. Do’s and Don’ts of Alpha

IX. Q Learning Equation

X. Optimal Value for number of Episodes

XI. When to Stop Training

b. Markov Decision Process

i. Agent-environment interaction

ii. Goals

iii. Returns

iv. Episodes

v. Value functions

vi. Optimization of policy

vii. Optimization of the value function

viii. Approximations

ix. Exercises and Thoughts

c. Q-Learning

i. Intro to QL

ii. Equation Explanation

iii. Implementation using Python

iv. Off-Policy Learning

d. SARSA

i. Intro to SARSA

ii. State, Action, Reward, State, Action

iii. Equation Explanation

iv. Implementation using Python

v. On-Policy Learning

e. Q-Learning vs. SARSA

i. Difference in Equation

ii. Difference in Implementation

iii. Pros and Cons

iv. When to use SARSA

v. When to use Q Learning

vi. Quiz/Solution

4. Mini Project Using the Above Concepts (Frozen Lake)

a. Intro to GYM

b. Gym Environment

c. Intro to Frozen Lake Game

d. Rules

e. Implementation using Python

f. Agent Evaluation

g. Conclusion

5. Deep Learning/Neural Networks

a. Deep Learning Framework

i. Intro to Pytorch

ii. Why Pytorch?

iii. Installation

iv. Tensors

v. Auto Differentiation

vi. Pytorch Practice

b. Architecture of DNN

i. Why DNN?

ii. Intro to DNN

iii. Perceptron

iv. Architecture

v. Feed Forward

vi. Quiz/Solution

vii. Activation Function

viii. Loss Function

ix. Gradient Descent

x. Weight Initialization

xi. Quiz/Solution

xii. Learning Rate

xiii. Batch Normalization

xiv. Optimizations

xv. Dropout

xvi. Early Stopping

c. Implementing DNN for CIFAR Using Python

6. Deep RL / Deep Q Network (DQN)

a. Getting to DQN

i. Intro to Deep Q Network

ii. Need of DQN

iii. Basic Concepts

iv. How DQN is related to DNN

v. Replay Memory

vi. Epsilon Greedy Strategy

vii. Quiz/Solution

viii. Policy Network

ix. Target Network

x. Weights Sharing/Target update

xi. Hyper-parameters

b. Implementing DQN

i. DQN Project – Cart and Pole using Pytorch

ii. Moving Averages

iii. Visualizing the agent

iv. Performance Evaluation

7. Car Racing Project

a. Intro to game

b. Implementation using DQN

8. Trading Project

a. Stable Baseline

b. Trading Bot using DQN

9. Interview Preparation

Successful completion of this course will enable you to

● Relate the concepts and practical applications of Reinforcement and Deep Reinforcement Learning with real-world problems.

● Apply for the jobs related to Reinforcement and Deep Reinforcement Learning.

● Work as a freelancer for jobs related to Reinforcement and Deep Reinforcement Learning.

● Implement any project that requires Reinforcement and Deep Reinforcement Learning knowledge from scratch.

● Extend or improve the implementation of any other project for performance improvement.

● Know the theory and practical aspects of Reinforcement and Deep Reinforcement Learning.

Who this course is for

● Beginners who know absolutely nothing about Reinforcement and Deep Reinforcement Learning.

● People who want to develop intelligent solutions.

● People who love to learn the theoretical concepts first before implementing them using Python.

● People who want to learn PySpark along with its implementation in realistic projects.

● Machine Learning or Deep Learning Lovers.

● Anyone interested in Artificial Intelligence.

Who this course is for
● Beginners who know absolutely nothing about Reinforcement and Deep Reinforcement Learning.
● People who want to develop intelligent solutions.
● People who love to learn the theoretical concepts first before implementing them using Python.