This is the code for "How to Do Win Slot Machines - Intro to Deep Learning #13' by Siraj Raval on YouTube
The coding challenge for this video is to use multiple slot machines instead of one. This way, state is taken into account. See this article for more info on this. Bonus points given for applying the code to a real world use case. You'll learn more about how policy and value functions are related in reinforcement learning by doing this exercise.
This is the code for this video on Youtube by Siraj Raval as part of the Udacity Intro to Deep Learning nanodegree. This code implements a technique called policy gradients to solve the multi-armed bandit problem. We use only 1 slot machine with 4 arms in this code and use an epsilon greedy policy to help select the best actions. Our agent is a simple 1 layer neural network built in tensorflow.
- tensorflow (https://www.tensorflow.org/install/)
- numpy
Run jupyter notebook in the main directory of this repository in terminal to see the code pop up in your browser.
Install jupyter here if you haven't already.
The credits for this code go to awjuliani. I've merely added a wrapper to get people started.