Hi! You have found the code of Deep Learning Group 6 for Assignment 2, task 2. In this assignment, we implemented NeuMF using the Microsoft recommenders package.
To install all packages that our code is dependent on:
pip install -r /path/to/requirements.txtWe used Python 3.7, as this is the only version guaranteed to work with the Microsoft recommenders package.
When you have installed the correct version of Microsoft recommenders, you are not there yet. There is still legacy code regarding the NCF part of this package in the package you download from pip. What you now do is you go to: https://github.com/microsoft/recommenders/tree/main/recommenders/models/ncf and replace the files dataset.py and ncf_singlenode.py with the version you find on this page. Furthermore, in order for the recommenders package to work, make sure you have installed Microsoft C++ Build Tools and have followed all the steps mentioned in their README file in order for the package to work as intented.
Create a directory named Data in your project folder and zip the contents of: https://files.grouplens.org/datasets/movielens/ml-100k.zip in this folder. Specifically you need the file named u.data.
To run both models:
python main2.pySet the variable explore_data in main2.py:
explore_data = TrueSet the variable grid_search in main2.py:
grid_search = TrueFurthermore, in the top of the main2.py file you can set the following settings: top_k (top k items to recommend), seed (random seed), n_epochs, batch_size, learning_rate, layer_sizes_list (architecture of MLP), n_factors_list (which number of factors for GMF embedding).