Chapter 6. Alternative optimization methods: Evolutionary algorithms

 

This chapter covers

  • Evolution algorithms for solving optimization problems
  • Pros and cons of evolutionary approaches versus previous algorithms
  • Solving the CartPole game without backpropagation
  • Why evolutionary strategies can scale better than other algorithms

Neural networks were loosely inspired by real biological brains, and convolutional neural networks were also inspired by the biological mechanism of vision. There is a long tradition of advances in technology and engineering being motivated by biological organisms. Nature, through the process of evolution by natural selection, has solved many problems elegantly and efficiently. Naturally, people wondered whether evolution itself could be borrowed and implemented on a computer to generate solutions to problems. As you will see, we can indeed harness evolution to solve problems, and it works surprisingly well and is relatively easy to implement.

6.1. A different approach to reinforcement learning

6.2. Reinforcement learning with evolution strategies

6.3. A genetic algorithm for CartPole

6.4. Pros and cons of evolutionary algorithms

6.5. Evolutionary algorithms as a scalable alternative

Summary

sitemap