Information Theoretic Counterfactual Learning from Missing-Not-At-Random Feedback

Authors: Zifeng Wang, Xi Chen, Rui Wen, Shao-Lun Huang, Ercan Kuruoglu, Yefeng Zheng

NeurIPS 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Empirical evaluation on real-world datasets shows that our CVIB significantly enhances both shallow and deep models
Researcher Affiliation Collaboration Zifeng Wang1,Xi Chen2,Rui Wen2,Shao-Lun Huang1,Ercan E. Kuruoglu1,3,Yefeng Zheng2 1Tsinghua-Berkeley Shenzhen Institute, Tsinghua University 2Jarvis Lab, Tencent 3Institute of Science and Technologies of Information, CNR, Pisa, Italy
Pseudocode Yes Algorithm 1 Counterfactual Learning with CVIB in MNAR Data for Recommendation.
Open Source Code Yes Code is available at https://github.com/Ryan Wang Zf/CVIB-Rec.
Open Datasets Yes Yahoo! R3 Dataset [25]. This is a user-song rating dataset... Coat Shopping Dataset [30]. This dataset consists of...
Dataset Splits Yes We randomly draw 30% data from the training set for validation, on which we apply a grid search for hyperparameters to pick the best configuration.
Hardware Specification No The paper does not specify the exact GPU models, CPU models, or other detailed hardware specifications used for running the experiments.
Software Dependencies No The paper mentions 'Py Torch [26]' but does not provide a specific version number for it or other software dependencies.
Experiment Setup Yes For both the MF and NCF, we fix the embedding size of both users and items to be 4... Adam [20] is utilized as the optimizer for fast convergence during training, with its learning rate in {0.1, 0.05, 0.01, 0.005, 0.001}, weight decay in {10 3, 10 4, 10 5}, and batch size in {128, 256, 512, 1024, 2048}. For NCF, we set an additional hidden layer with width 8. Specifically for CVIB, we set the hyperparameters α {2, 1, 0.5, 0.1}, and γ {1, 0.1, 10 2, 10 3}.