English | MP4 | AVC 1920×1080 | AAC 48KHz 2ch | 1h 23m | 306 MB
Solve real-world problems by employing reinforcement learning techniques with Java
There are problems in data science and the ML world that cannot be solved with supervised or unsupervised learning. When the standard ML engineer’s toolkit is not enough, there is a new approach you can learn and use: reinforcement learning.
This course focuses on key reinforcement learning techniques and algorithms in the Java ecosystem. Each section covers RL concepts and solves real-world problems. You will learn to solve challenging problems such as creating bots, decision-making, random cliff walking, and more. Then you will also cover deep reinforcement learning and learn how you can add a deep neural network with DeepLearning4J in your RL algorithm.
By the end of this course, you’ll be ready to tackle reinforcement learning problems and leverage the most powerful Java DL libraries to create your reinforcement learning algorithms.
Use reinforcement learning with DL4J and RL4J to solve problems with high accuracy
Learn how to use the ND4J and RL4J libraries with external libraries such as Malmo to abstract complex algorithms and make them easy to use
Implement q-learning, Markov Decision Processes (MDPs), dynamic programming, and other reinforcement techniques to solve real-world problems
Learn
- Leverage ND4J with RL4J for reinforcement learning
- Use Markov Decision Processes to solve the cart-pole problem
- Use QLConfiguration to configure your reinforcement learning algorithms
- Leverage dynamic programming to solve the cliff walking problem
- Use Q-learning for stock prediction
- Solve problems with the Asynchronous Advantage Actor-Critic technique
- Use RL4J with external libraries to speed up your reinforcement learning models
Table of Contents
Deep Dive into Reinforcement Learning with DL4J RL4J
1 The Course Overview
2 Main Principles of Reinforcement Learning
3 Adding DL4J with RL4J to Our Project
4 Best Use Cases of Reinforcement Learning
5 Configuring Reinforcement Learning Model with QLearning.QLConfiguration
Solving Cartpole with Markov Decision Processes (MDPs)
6 Understanding Cartpole Problem
7 Leveraging Markov Chain in Our Cartpole Solution
8 Using QLConfiguration to Configure Our Model
9 Using GymEnv Library from RL4J to Simulate Solution
10 Running Cartpole and Validating Results
Using Project Malmo Reinforcement Learning Leveraging Dynamic Programming
11 Adding Malmo Library to Our RL4J Project
12 Analyzing Possible Scenarios That Our Program Can Solve
13 Loading Cliff Walking Simulation
14 Configuring RL4J Algorithm for Cliff Walking Problem
15 Starting QLearningDiscreteDense and Saving Results
Creating Decision Process for Stock Prediction with Rewards Using Q-Learning
16 Understanding Stock Prediction Problem
17 Creating Configuration for Stock Prediction Learning
18 Leveraging QLearningDiscreteDense from RL4J API
19 Performing Stock Prediction Training and Validating Results
Leveraging Monte Carlo Tree Searches and Temporal Difference (TD) in RL
20 Understanding Asynchronous Advantage Actor-Critic Technique(A3C)
21 Setting Up A3C Learning Environment
22 Configuring Reinforcement Learning Program Using A3C Configuration
23 Using A3C Technique with ActorCriticFactorySeparateStdDense
24 Starting Program and Gathering Results
Resolve the captcha to access the links!