wangyongjie-ntu / GUM

The official code for the paper "Summarizing User-Item Matrix By Group Utility Maximization"

GUM

GUM(Group Utility Maximization) is the official repository for the paper "Summarizing User-Item Matrix By Group Utility Maximization"(in submission).

Requirements

pip install captum, numpy, pandas, pytorch, sklearn

Getting Start

The listed four folders.

• preprocess/titanic_importance.py specifies how to train the model and obtain the feature importance matrix. The pretrained model is provided too.
• data/ folder consists of the raw titanic dataset, feature importance matrix of titanic and netflix-200 data processed by ours. Limited by the GitHub capacity, users can download the netflix-prize and Movielens datasets from their official links netflix-prize-data and Movielens.
• algorithm/ folder contains the CELF (accelerated greedy), stochastic greedy, k-max, brute_force, and baselines in our paper.
• run/ folder provides examples to run the algorithms on different datasets.

Example

The such code snippet shows how to obtain the group summarization on Titanic dataset. You can reproduce the experiments by replacing the dataset and setting the parameters $k$ and $l$ of our examplar codes.

cd run;
python titanic_run.py

Cite Us