hyperrs

This project follows BSD 3-Clause License.

Source Code for the paper "HyperRS: Hypernetwork-based Recommender System for the User Cold- Start Problem"

Dependencies:

tqdm tensorflow 2.8.0 numpy

Requirement:

At least 10GB GPU memory. At least 32GB memory.

1. Download the preprocessed data from (this will be provided on request.)
2. Unzip the data.tar.gz. Move all files in the same directory with trainer.py
3. Modify DATASET in dataset_info.json to choose ml-10m (Movielens), tokyoTV (Tokyo TV), Book (BookCrossing)
4. Use python trainer.py -m [CONFIG_FILE_NAME] to train the dataset. [CONFIG_FILE_NAME]: HyperRS.ml-10m.C-W for Movielens, Cold User Warm Item.

Change the use_existing_items in the [CONFIG_FILE_NAME] to false to test the Cold User Cold Item setting.