GuidedRec

This is our implementation for the SIGIR 2021 paper:

Rashed, Ahmed, Josif Grabocka, and Lars Schmidt-Thieme. "A Guided Learning Approach for Item Recommendation via Surrogate Loss Learning (SIGIR). 2021.

Enviroment for GuidedRec, Surrogate Loss and Logloss

* pandas==1.0.3
* tensorflow==1.14.0
* matplotlib==3.1.3
* numpy==1.18.1
* six==1.14.0
* scikit_learn==0.23.1

Commands

GraphRec Model

with logloss

• python GraphRec.py 0 45

with GuidedRec

• python GraphRec.py 1 45

surrogate Only

• python GraphRecSurrogate.py 1 45

NueMF Model

with logloss

• python NueMF.py 0 45

Enviroment for TFRanking Losses

* numpy==1.18.1
* six==1.14.0
* matplotlib==3.1.3
* tensorflow==2.3.0
* pandas==1.0.3
* scikit_learn==0.23.1
* tensorflow_addons==0.10.0
* tensorflow_ranking==0.3.0

Commands

GraphRec Model

with gumbel_approx_ndcg_loss

• python GraphRecTFRank.py 0 45 "'gumbel_approx_ndcg_loss'"

with approx_ndcg_loss

• python GraphRecTFRank.py 0 45 "'approx_ndcg_loss'"

with list_mle_loss

• python GraphRecTFRank.py 0 45 "'list_mle_loss'"

with softmax_loss

• python GraphRecTFRank.py 0 45 "'softmax_loss'"

with pairwise_logistic_loss

• python GraphRecTFRank.py 0 45 "'pairwise_logistic_loss'"

with neural_sort_cross_entropy_loss

• python GraphRecTFRank.py 0 45 "'neural_sort_cross_entropy_loss'"

Note: All scripts require a gpu. Please change the device <DEVICE = "/gpu:0"> to cpu if you have no gpu