This repo is full of code for CS 7641 - Machine Learning at Georgia Tech.
A huge thanks to jontay (https://github.com/JonathanTay) for sharing his code. Much of the code contained in this repo is based off of his work.
Yup, we were encouraged to steal code. All the code. It was fine. Only the analysis mattered.
For dated support of this claim, see https://gist.github.com/cmaron/46f0992d42be87380c208086eec9797f
For even more dated support, we were asked to submit links to code rather than the actual code when submitting assignments partway through the semester.
Assignments 1, 3, and 4 require python (specifically python 3). Assignment 2 is a bit... unique... in that it uses Jython. Check the readmes for the specific assignments for more details.
If a python virtual environment has been setup for the project, a simple pip install -r requirements.txt
should take care of the required packages.
Each assignment folder has its own run_experiment.py
that will do most of the work for you. The big exception is assignment 2. Assignment 2, at least in Fall of 2018, was due soon after the midterm which was soon after the first assignment. These assignments take a while so I didn't put a ton of effort into doing anything fancy for assignment 2. Not to say any of this is fancy, obviously.
Running python run_experiment.py -h
should provide a list of options for what you can do.
For the most part it is simple to run a given set of experiments based on a specific algorithm. One flag to consider always including is --threads
with a value of -1
. This will speed up execution in some cases but also might use all available cores.
The --verbose
flag can be helpful to view data about a given dataset or MDP.
For assignments 3 and 4 plotting data is a separate step from generation. For those assignments the --plot
flag should be used once data is generated
Each assignment folder should have its own readme with anything specific to not for that assignment.
Good question.
Ok fine, I did ok on the assignments (as in no grade less than 90). For the grades that were not 100 the feedback did not mention missing charts or values, so I'm confident this code does not miss anything major in that regard.
That said, this is based off of the Fall 2018 semester. Things can change and you should always pay attention to announcements and go to office hours to be certain of the specifics.
Feel free to open an issue for things that are flat out broken (or even better open a PR) and I can take a look.
That said, caveat emptor applies.