marshallcn / rechash Goto Github PK

###############################################
Parameters
###############################################

--dataset', type=string, default='', help='Name of the dataset, movielens: using MovieLens dataset, amazon: using Amazon dataset')
--mode, type=string, default='user_based', help='user_based: using user-based CF, item_based: using item-based CF')
--emb_method, type=int, default=0, help='0: using user-item interaction matrix (RH), 1: using metapath2vec for generating user/item embeddings (ME) for CF-KNN')
--blocks_metapath, type=string, default='', help='User/Item blocks meta-path')

# Parameters for KNN method
--n_neighbors, type=int, default=10, help='Number of user/item neighbors for CF')


# Paramters for metapath2vec (for ME approach only)
--kg_file, type=string, default='relations_vbpr_100.json', help='KG filename')
--n_walk, type=int, default=100, help='Maximum number of walks per starting node for metapath2vec (for ME only)')
--embsize', type=int, default=300, help='Node embedding size for metapath2vec (for ME only)')
--metapath, type=string, default='', help='Meta-path for generating node embeddings with metapath2vec (for ME only)')
--prs, type=string, default='[0,0]', help='Probabilities of going to visual node type when using visually-annotated meta-paths (for ME only)')


# Evaluation
--eval, type=bool, default=True, help='True: perform evaluation, False: do not perform evaluation')
--listK, type=string, default='[1, 5, 10, 50, 100]', help='List of K for top-K recommendations')





###############################################
Example of how to run the code:
###############################################

Before running the code, plese run this follwing command:
>>> chmod +x code_metapath2vec/metapath2vec

Using user-based KNN method and RH_UP_6bit_1r_BU user blocks on movielens dataset (The user embeddings are obtained from the user-item interaction matrix)
>>> python3 main.py --dataset movielens --mode user_based --blocks_metapath RH_UP_6bit_1r_BU --emb_method 0

Using item-based KNN method and Ph(P)_BP item blocks on movielens dataset (The item embeddings are obtained from the user-item interaction matrix)
>>> python3 main.py --dataset movielens --mode item_based --blocks_metapath Ph\(P\)_BP --emb_method 0

Using user-based KNN method and RH_UP_6bit_1r_BU user blocks on movielens dataset (The user embeddings are generated by metapath2vec based on meta-path UPTP) 
python3 main.py --dataset amazon --mode user_based --blocks_metapath RH_UP_6bit_1r_BU --emb_method 1 --metapath UPTP