Hi Jeff! Thank you very much for comenting. Yes, there's a Github repository with the code I used for the project https://github.com/diegomontoyas/DeepLearning-PacMan.

I agree. Minimizing error is key to generate better conclusions, which includes repeating experiments and averaging results. For instance, I think one of the ways the project could be improved is by doing smaller experiments, so as to minimize the number of variables that can affect the results. So, for example, instead of testing three things at once in one project, just choosing one and doing a more in-depth research on that one.