Projected-Hamming-Dissimilarity

Christian Hansen*, Casper Hansen*, Jakob Grue Simonsen, Christina Lioma. Projected Hamming Dissimilarity for Bit-Level Importance Coding in Collaborative Filtering. in Proceedings of the Web Conference 2021 (WWW'21).
* denotes equal contribution.

The structure of the code is as described below. The model is implemented in tensorflow 1.12 using python 3.6.

data directory

• This contains the txt2mat.py script that converts the raw data (provided via the links within the paper) to .mat files.
• Use this link to download the datasets and splits used exactly in our paper: https://www.dropbox.com/s/v69u3l9e897egqp/data.zip?dl=0

code directory

• This contains the implementation of our model.
• main.py is the main file, which runs and evaluates the model. Check the argparse arguments on line 97 for how to run on different datasets and different configurations.
• model.py contains the model implementation.
• tf_dataset_maker.py transforms .mat files into tfrecords used by the model.