Code for the SIGIR20 paper -- Recommendation for New Users and New Items via Randomized Training and Mixture-of-Experts Transformation

We used three datasets (CiteULike, LastFM, and XING) in this work, which are stored in 'data' folder. Details about these datasets can be found in the paper and also the original papers cited in our paper.

python 3
tensorflow 1.14.0 numpy sklearn pandas

Run main_CiteULike.py to run the model Heater on CiteULike data; Run main_LastFM.py to run the model Heater on LastFM data; Run main_XING.py to run the model Heater on XING data;

Our code is based on the implementation of DropoutNet (https://github.com/layer6ai-labs/DropoutNet).