Auxiliary_enhanced_ALS

The implementation of AALS

Auxiliary-enhanced eALS performs well by integrating users' multiple auxiliary data as the intermediate feedback. This is our official implementation for the paper:

Jingtao Ding, Guanghui Yu, Yong Li, Xiangnan He, Depeng Jin and Jiajie Yu, Improving Implicit Recommender Systems with Auxiliary Data, To appear in ACM Transactions on Information Systems (TOIS).

If you use the codes, please cite our paper. Thanks!

Requirements

JAVA

Currently, for baselines, the eALS, BPR, MC-BPR, RankALS are provided.

The implementation of proposed pointwise and pairwise AALS methods are capable of incorporating one or two types of auxiliary data.

Quick Start

run Pair2AALS

java -jar <name_of_jar>.jar main_MF <path_of_primary_file> Pair2AALS <value_s0> False True <number_factors> <number_iterations> <reg_value> 0 <path_of_auxiliary1_file> <value_c0> 0 <value_gamma1> <value_gamma2> <path_of_auxiliary2_file> <value_c1> <value_gamma3> <flag_eval_topk> <validation_flag>

run Point2AALS

java -jar <name_of_jar>.jar main_MF <path_of_primary_file> Point2AALS <value_s0> False True <number_factors> <number_iterations> <reg_value> 0 <path_of_auxiliary1_file> <value_c0> 0 <value_r1> 0 <path_of_auxiliary2_file> <value_c1> <value_r2> <flag_eval_topk> <validation_flag>

run Pair1AALS

java -jar <name_of_jar>.jar main_MF <path_of_primary_file> Pair1AALS <value_s0> False True <number_factors> <number_iterations> <reg_value> 0 <path_of_auxiliary1_file> <value_c0> 0 <value_gamma1> <value_gamma2> <validation_flag>

run Point1AALS

java -jar <name_of_jar>.jar main_MF <path_of_primary_file> Point1AALS <value_s0> False True <number_factors> <number_iterations> <reg_value> 0 <path_of_auxiliary1_file> <value_c0> 0 <value_r1> 0 <validation_flag>

run eALS

java -jar <name_of_jar>.jar main_MF <path_of_primary_file> fastals <value_s0> False True <number_factors> <number_iterations> <reg_value> 0

run RankingALS

java -jar <name_of_jar>.jar main_MF <path_of_primary_file> rankingals 0 False True <number_factors> <number_iterations> <reg_value> 0 <path_of_auxiliary1_file> <value_r1> <path_of_auxiliary2_file> <value_r2> <validation_flag>

run BPR

java -jar <name_of_jar>.jar main_MF <path_of_primary_file> bpr <learning_rate> False True <number_factors> <number_iterations> <reg_value>

run MC-BPR

java -jar <name_of_jar>.jar mfbpr_pos 0 0 <learning_rate> False True <number_factors> <number_iterations> <reg_value> 0 <path_of_primary_file> <path_of_auxiliary1_file> false <value_beta1> <value_beta2>