Genetic Algorithm for Reinforcement Learning : Python implementation
Most beginners in Machine Learning start with learning Supervised Learning techniques such as classification and regression. However, one of the most important paradigms in Machine Learning is Reinforcement Learning (RL) which is able to tackle many challenging tasks. It is an aspect of Machine learning where an agent learns to behave in an environment, by performing certain actions and observing the rewards(results) which it gets from those actions.
In Reinforcement Learning, we give the machines a few inputs and actions, and then, reward them based on the output. Reward maximization is the end goal. It is just like a little baby who knows nothing at first is left alone in an environment and then after a little understanding tends to know things around itself.
Here, we are just going to build an algorithm based on the genetic mutation of a population when attacked by a virus. In the first generation of our population only a few fittest ones will be able to survive, whereas as the generations will pass, the new generations will be much stronger against the virus than their ancestors. It is a basic algorithm which just gives an idea of how these things work. Anyone with the basic knowledge of python and some libraries like numpy, matplotlib, etc can easily understand this code. This is just for the introduction and to provide the surface level knowledge about Reinforcement Learning.
Genetic Algorithm –
- numpy : we’ll be using numpy arrays and other basic calculation functionalities from this library
- matplotlib : we’ll be using matplotlib.pyplot functionality in order to plot the graphs for the visual understanding of the algorithm.
In this program, we’ll define 3 main functions in order to generate the next generation of the population which is genetically more powerful than the previous ones.
The three main functions used are:
populate: This is used to generate the population and then appending it to a list. This function accepts the attributes like the number of features in the population and the size of it.
reduction: This function is used to reduce the size of the population and allows only the 100 most fittest ones to survive. These fit ones will be the ones used to generate the next generation of the population.
cross: This function is used for the process of cross-breeding between the ones that are left in order to generate a whole new generation of the population which will be much more immune towards the virus.
Below is the implementation –
1) For generation 1, when gens=0
2) For generation 2, when gens=1
3) For generation 3, when gens=2