author: Patrick Buhagiar
date: 11/09/2018
- Install python 3.6+
- Using
pip3, install:numpypandastensorflow
All the experiments are in the src folder. Each expirment is named after the requirements in Chapter 3. To run the code, execute in command promt or console.
python3 h1.py
Replace with relevant experiment. Each experiment outputs two .csv files, one for accuracy and another for parameters. These files can the be read by the result_plotter.py file, after changing the appropriate parameters and file name in the file.
Stage 1 of h2 and h3 persist models under the h2_models and h3_models. These folders already consist of the models created in this experiment, thus stage 1 can be skipped.
Finally, the unparametric_test.py is used to output the predictions for each model. these predictions, found under the prediction folder, can be used for the Kruskall-Wallis test and Wilcoxon test.
These experiments were run on Google Cloud, with VMs set up with 8-20 cores, thus they were meant to run on several threads and cores.