Sparse-LInear-Models (SLIM)

SLIM model for top_n recommendation

This is a C++ implementation of the top_n recommendation algorithm SLIM from the paper

Ning, Xia, and George Karypis. "Slim: Sparse linear methods for top-n recommender systems." 2011 11th IEEE International Conference on Data Mining. IEEE, 2011.

Requirements

• gcc
• OpenMP

Usage

1. Type make to compile the code and generate the excutable
2. Run ./slim data_file l1 l2 top_n num_threads eps

Args

• data_file: path to the dataset
• l1: hyperparameter for l1 regularization which controls the sparsity of the model
• l2: hyperparameter for l2 regularization
• top_n: length of recommended lists
• num_threads: number of threads available to accelerate the training process
• eps: number to control the convergence of the model

Data format

The program treats each interaction as implicit feedback. The dataset shoud have three columns and be formated as

user_id, item_id, timestamp

A sample dataset amazon_game_5.csv is uploaded.