EDM591_Hyperparameter

• Final PPT can be found here: https://tiny.cc/edm17ppt
• Final Report can be found here: https://tiny.cc/edm17report
• Github Repository can be found here: https://github.com/amritbhanu/EDM591_Hyperparameter

Installation:

• Run 'pip install -r requirements.txt' to get the required python packages.
• Need python2.7

Directory Structure:

• Data folder contains raw data as well as preprocessed data. You will need raw folder and datset2 folder inside. dataset1, dataset2, dataset3 csvs will automatically be generated.
• Dump folder contains the results dump from running our scripts. So that results can be generated quickly.
• Results folder contain all our graphs and results for the report and ppt. They will automatically be generated our running our scripts.
• Src directory contains all our scripts.

Src scripts:

• DE.py is the generalised code/class of our DE. We are planning to publish this DE as a python package.
• demos.py is the code which runs our main script by calling its function and parameters as argument.
• main.py is the main code which runs our tuned results and generates dump.
• ML.py is the generalised code of all our Machine learning implementations.
• Preprocess_dataset1_3.py is the preprocess script for dataset1 and dataset3.
• preprocessing_dataset2.py is the preprocess script for dataset2.
• read_pickle.py reads all the results dump from dump folder and generates graph in results folder.
• sk.py is the code for our statistical test which is scottknot.
• untuned.py is another main code which runs our untuned results and generates dump.

How to run scripts:

Go into src folder and run in the sequential order, how we mentioned below. 1) 'python preprocessing_dataset2.py' 2) 'python Preprocess_dataset1_3.py' 3) 'python untuned.py _test dataset1' : this will generate dataset1_untuned.pickle in dump folder 4) 'python untuned.py _test dataset2' : this will generate dataset2_untuned.pickle in dump folder 5) 'python untuned.py _test dataset3' : this will generate dataset3_untuned.pickle in dump folder 6) Now to run these scripts you will need High Performance computing (HPC) servers since it will 4-8 hours to end each script. If it cant be run, we have provided the dump of our results. Directly jump to step 7.

• 'python main.py _test dataset1' : this will generate dataset1.pickle and dataset1_late.pickle in dump folder
• 'python main.py _test dataset2' : this will generate dataset2.pickle and dataset2_late.pickle in dump folder
• 'python main.py _test dataset3' : this will generate dataset3.pickle and dataset3_late.pickle in dump folder 7) 'python read_pickle.py' : will generate graphs in results folder.