The file document.pdf contains the most recent version, built from the latex files and references the source code in places.

Selected images (including those used in the document) are contained in the /img directory

Code is found in the python modules, as well as jupyter notebooks where indicated (or linked in the main document).

The document was initially based on the Template by Michael Kohlegger.

Contained in the /lst directory

Any Data used in the Jupyter Notebooks and the source codes is contained in the /data directory, original sources are listed in the main document.

Pretrained neural networks are stored in the /models directory as .h5 files

Strict requirements are listed in the requirements.txt - Python version used was python 3.9.7 on arch linux

Since the installation of tensorflow is notoriously finnicky and this was written on arch linux with tensorflow installed from the AUR, your mileage may vary.

Some sample training logs can be found in the /custom_logs directory - to be viewed using TensorBoard

found in the ./notebooks folder, contain all the boosting and shrinking experiments

note: in order to run the jupyter notebooks, it is recommended to run jupyter in the top directory of this repository, since most import paths are relative