Removing Hidden Confounding in Recommendation: A Unified Multi-Task Learning Approach
Authors: Haoxuan Li, Kunhan Wu, Chunyuan Zheng, Yanghao Xiao, Hao Wang, Zhi Geng, Fuli Feng, Xiangnan He, Peng Wu
NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We conduct extensive experiments on three publicly available benchmark datasets containing a fully exposed large-scale industrial dataset, validating the effectiveness of the proposed methods in removing hidden confounding. |
| Researcher Affiliation | Academia | Haoxuan Li1, Kunhan Wu2, Chunyuan Zheng3, Yanghao Xiao4, Hao Wang5, Zhi Geng6, Fuli Feng7, Xiangnan He7, Peng Wu6, 1Peking University 2Carnegie Mellon University 3University of California, San Diego 4University of Chinese Academy of Sciences 5Zhejiang University 6Beijing Technology and Business University 7University of Science and Technology of China |
| Pseudocode | No | The paper does not contain structured pseudocode or algorithm blocks, nor does it refer to any figure as an algorithm or pseudocode. |
| Open Source Code | No | The paper does not provide concrete access to source code for the methodology described, nor does it include an explicit code release statement or a repository link. |
| Open Datasets | Yes | Following the previous studies [4, 42, 52], we use three real-world datasets: COAT2, YAHOO! R33, and KUAIREC4 [9], for evaluating the debiasing performance of the proposed methods, where KUAIREC is a public large-scale industrial dataset. 2https://www.cs.cornell.edu/~schnabts/mnar/ 3http://webscope.sandbox.yahoo.com/ 4https://github.com/chongminggao/KuaiRec |
| Dataset Splits | No | In addition, we randomly split 5% of unbiased data from the test set to train models for all methods that require unbiased data5. The paper does not provide specific dataset split information (exact percentages, sample counts, citations to predefined splits, or detailed splitting methodology) for training, validation, and testing of the primary models. |
| Hardware Specification | Yes | For all experiments, we use Tesla T4 GPU as the computational resource. |
| Software Dependencies | No | All the experiments are implemented on Pytorch with Adam as the optimizer. The paper mentions software names but does not provide specific version numbers for key components like Pytorch. |
| Experiment Setup | Yes | We tune learning rate in {0.0001, 0.0005, 0.001, 0.005, 0.01, 0.05}, weight decay in {0, 1e 6, . . . , 1e 1, 1}. For our methods, we tune α in {0.1, 0.5, 1}, β in {0.1, 0.5, 1, 5, 10}, and γ in {0.001, 0.005, 0.01, 0.05, 0.1}. |