Ducho meets Elliot: Large-scale Benchmarks for Multimodal Recommendation

Official repository for the paper Ducho meets Elliot: Large-scale Benchmarks for Multimodal Recommendation. The codebase was developed and tested on Ubuntu 22.04 LTS; however, the experiments can be executed on other operating systems with the necessary adjustments to environment variables, activation of Python environments, or configuration of additional utilities.

Installation

Clone the repo

This repository integrates the Ducho and Elliot frameworks. The Ducho framework is hosted in a separate GitHub repository. To properly clone the Ducho-meets-Elliot repository, please use the following command:

git clone --recursive https://github.com/sisinflab/Ducho-meets-Elliot.git

Build the environment

We recommend using Conda to create virtual environments with all the required libraries. For instructions on installing Conda, please refer to the official documentation.

To create the environment, run the following command:

conda env create -f ducho_env.yml

After the installation is complete, activate the environment by running the following command:

conda activate ducho_env

Download and pre-process the datasets

To download the datasets used in our experiments, please go into the Ducho submodule.

cd Ducho

Then, run the following command for the desired dataset:

python3 ./demos/demo_dataset_name/prepare_dataset.py

Running the experiments

After the preprocessing of the selected dataset is complete, ensure you are in the project's root directory and run the following command to execute the experiments. Be careful to replace dataset_name and batch_size with the desired ones.

bash run_experiments.sh dataset_name batch_size

The Team

Currently, this repository is mantained by:

• Matteo Attimonelli (matteo.attimonelli@poliba.it)
• Danilo Danese (danilo.danese@poliba.it)
• Angela Di Fazio (angela.difazio@poliba.it)
• Daniele Malitesta (daniele.malitesta@poliba.it)
• Claudio Pomo (claudio.pomo@poliba.it)
• Tommaso Di Noia (tommaso.dinoia@poliba.it)