Book Image

Hands-On Genetic Algorithms with Python

By : Eyal Wirsansky
Book Image

Hands-On Genetic Algorithms with Python

By: Eyal Wirsansky

Overview of this book

Genetic algorithms are a family of search, optimization, and learning algorithms inspired by the principles of natural evolution. By imitating the evolutionary process, genetic algorithms can overcome hurdles encountered in traditional search algorithms and provide high-quality solutions for a variety of problems. This book will help you get to grips with a powerful yet simple approach to applying genetic algorithms to a wide range of tasks using Python, covering the latest developments in artificial intelligence. After introducing you to genetic algorithms and their principles of operation, you'll understand how they differ from traditional algorithms and what types of problems they can solve. You'll then discover how they can be applied to search and optimization problems, such as planning, scheduling, gaming, and analytics. As you advance, you'll also learn how to use genetic algorithms to improve your machine learning and deep learning models, solve reinforcement learning tasks, and perform image reconstruction. Finally, you'll cover several related technologies that can open up new possibilities for future applications. By the end of this book, you'll have hands-on experience of applying genetic algorithms in artificial intelligence as well as in numerous other domains.
Table of Contents (18 chapters)
1
Section 1: The Basics of Genetic Algorithms
4
Section 2: Solving Problems with Genetic Algorithms
9
Section 3: Artificial Intelligence Applications of Genetic Algorithms
14
Section 4: Related Technologies

Solving the CartPole environment

The CartPole-v1 environment simulates a balancing act of a pole, hinged at its bottom to a cart, which moves left and right along a track. Balancing the pole upright is carried out by applying to the cart one unit of force—to the right or to the left—at a time.

The pole, acting as a pendulum in this environment, starts upright within a small random angle, as shown in the following rendered output:

CartPole simulation—starting point

Our goal is to keep the pendulum from falling over to either side for as long as possible, that is, up to 500 time steps. For every time step that the pole remains upright, we get a reward of +1, so the maximum total reward is 500. The episode will end prematurely if one of the following occurs during the run:

  • The angle of the pole from the vertical position exceeds 15 degrees.
  • The cart's...