This folder contains all code and dataset related to the the thesis for MSc. in AI.
The folder structure is as follows:
-
data:
- dataset_1: Synthetic dataset trained on TVAE from real data (main artifact for the thesis)
- processed: processed dataset after running model/1_dataprep.ipynb
- dataset_2: Secondary dataset from https://www.kaggle.com/datasets/parisrohan/credit-score-classification
- raw: raw dataset as downloaded
- dataset_1: Synthetic dataset trained on TVAE from real data (main artifact for the thesis)
-
envs: virtual environment yaml file to run a replicated environment for running all code
-
imgs: images used during the thesis write up.
-
model: all code sits inside this directory
- logs: folder holds all logs from all agents experiment runs
- output: contains all serialized trained models
- utils:
- common.py : common code used across notebooks
- constants.py : constants to replicate sampled population
- doubledqn.py : personal implementation of double dqn extending stable baselines 3
- duelingdqn.py: personal implementation of dueling dqn extending stable baselines 3
- networks.py : personal implementation of different neural network arquitectures (feature extractors) for stable baselines 3
- riskenv.py : personal implementation of different environment for testing
- 1_dataprep.ipynb : all preprocessing of the raw data happens here.
- 2_modeling.ipynb : all modeling code for testing the agents and environment for dataset_1 (binary class)
- 3_modeling.ipynb : all modeling code for testing the agents and environment for dataset_2 (multi class Environment dataset)
** All code is commented for better legibility **