Simulation code for experiments in the paper: Multi-Aspects Joint Optimization in Recommendation Systems: A Mixture of Aspect-Explicit-Experts Architecture (submitted to KDD 2022).
Environment: Python 3.7.3
Versions of the other packages are summarized in requirements.txt
Please download the MovieLens 1M dataset at
Then unzip the package ml-1m.zip
and add the three datasets users.dat
, movies.dat
, and ratings.dat
to MAEE/SingleTask/data
and MAEE/Visualization_SingleTask/data
Please download the Census Income dataset at
-
https://archive.ics.uci.edu/ml/machine-learning-databases/census-income-mld/census-income.data.gz
-
https://archive.ics.uci.edu/ml/machine-learning-databases/census-income-mld/census-income.test.gz
Then add the two dataset census-income.data.gz
and census-income.test.gz
to MAEE/MultiTask/data
and MAEE/Visualization_MultiTask/data
-
Comparison of MAEE against DNN, MOE
-
Visualization of expert representations
-
Rationale analaysis of the masked attention mechanism
-
Comparison of MAEE against MMOE, PLE(CGC)
-
Visualization of expert representations