Jimmys Software Project

This week I continued on my learning journey of neural evolution of augmented topologies. I had some trouble learning about pickling, which I eventually understood and then realised I might not need it anyway. I found a very helpful set of video tutorials to build a neural network in python to play sonic the hedgehog, and I am hoping to follow that to help my understanding. Through that, I also discovered a vast open source library of interactable game environments for almost all retro games, called open ai retro, which is similar to open ai gym which I was using previously. I am still working on all the setup and package installation and hope to sort that out this week and get on with the tutorial, following that I will be attempting to implement a similar genetic algorithm to my own code, and develop a GUI which will hopefully end up to be my final project.

This week I continued to learn more about neural networks. However, I have been having a problem with installing box2d for python to simulate my bipedal walker. It seems like the only way I'm going to be able to do it is through conda, which I was trying to avoid since I had all the problems with it and the virtual env last time. Hopefully this week I will sort out the box2d issues and work on understanding more about deep learning.

This week I continued to learn more about neural networks and tensorflow. I followed a tutorial to help learn how to create an AI that taught itself how to balance a cart on a pole. The ai was reasonably competent, beating the required score. However, I'm looking to do more of a genetic algorithm. This coming week I will try to research more about genetic algorithms and NEAT AI to create a self learning ai that learns how to walk through genetic evolution and natural selection of multiple AI attempts.

Over the holidays I made progress with mostly the learning of neural networks, tensorflow and machine learning following tutorials to build things such as self-taught handwriting recognition software. I also resolved a lot of issues I had with packages and installation, I created a virtual environment through anaconda which allowed me to install packages I was unable to on my pc. However I then ran into the issue of some packages being unavailable on the virtual environment, and since I can only access one at a time, I didn't really have any ideas for a solution besides installing some sort of GitHub clone of the package into anaconda. Then during the holidays, my computer stopped working, and I got a replacement which led to me setting up all my packages again. This allowed me to install all of the libraries without having the previous issues or even the need for a virtual environment. Now that my packages are nicely setup and well structured I am able to focus more on the ac...