My first attempt at writing a neural network. Doens't get very far, but the paper I took inspiration from, atleast on the simpler NN, only achieved 256 whereas mine acheives 512 on some runs. If you have feedback on what I did wrong and such feel free to comment. Still very new to AI but would like to learn more. The code itself is a mess but I …
dhn045/2048AI
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
Folders and files
Name | Name | Last commit message | Last commit date | |
---|---|---|---|---|
Repository files navigation
Author: David Nguyen Date: 7/2/2018 2048 AI This is my first project applying a neural network to solve some kind of problem. The neural network used is a normal MLPNN that uses ideas in RL, reinforcement learning, to learn over time. I template code from deeplearningandneuralnetworks.com to learn and start and then heavily modified it with ideas from "Deep Reinforcement Learning for 2048", a paper written by Antoine Dediu and Jonothan Amar. The project was never made to be used by anyone but me so I don't plan on updating it, nor am I an expert an ML. Because of this, any ideas and concepts used in the code shouldn't be taken seriously. My goal was to meet the expectations that the MIT 2048 paper achieved with a random policy network with a CNN and RL, but it actually exceeded it with some different parameters. Introductory Online NN and DL Course that I used http://neuralnetworksanddeeplearning.com/chap1.html Deep Reinforcement Learning for 2048 http://www.mit.edu/~amarj/files/2048.pdf
About
My first attempt at writing a neural network. Doens't get very far, but the paper I took inspiration from, atleast on the simpler NN, only achieved 256 whereas mine acheives 512 on some runs. If you have feedback on what I did wrong and such feel free to comment. Still very new to AI but would like to learn more. The code itself is a mess but I …
Resources
Stars
Watchers
Forks
Releases
No releases published
Packages 0
No packages published