Collaborative Filtering

This project is part of the Computational Intelligence Lab course (2019) at ETH.

The project website can be found here.

Team: sasglentame

Name	Email
Sotiris Anagnostidis	[email protected]
Adamos Solomou	[email protected]
Ioannis Sachinoglou	[email protected]
----	[email protected]

Project structure

.
├── data                               # should contain files data_train.csv  sampleSubmission.csv
├── experiment_results                 # results from experiments
    ├── graphs                         # directory for saving graphs
    ├── preprocessed                   # directory for saving .csv
    ├── raw                            # directory for saving outputs
├── report                              
    ├── report.pdf                     # Final report
├── src 
    ├── experiments                    # experiment scripts
├── requirements.txt
└── README.md

Getting Started

To run on cluster:

# create environment
python -m venv "cil-2019"

# activate environment 
source cil-2019/bin/activate

# or use existing environment
# source $HOME/.local/bin/virtualenvwrapper.sh
# workon "cil-2019"

# install dependencies 
pip install --user -r requirements.txt

# load modules
module load python_gpu/3.6.4

To replicate final submission:

python main.py 

To replicate experiments:

cd src/experiments # You should be in the experiment directory for the experiments to run
python <experiment> 

To run cross-validation test:

python cross_validation.py --model <model> [--<parameters> <value>]

Valid models names are the following:

autoencoder, bsgd , svd_shrinkage, ncf

Documentation

Report