Towards a Theoretical Framework of Out-of-Distribution Generalization

Authors: Haotian Ye, Chuanlong Xie, Tianle Cai, Ruichen Li, Zhenguo Li, Liwei Wang

NeurIPS 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments on benchmark OOD datasets demonstrate that our model selection criterion has a significant advantage over baselines.
Researcher Affiliation Collaboration Haotian Ye Peking University Pazhou Lab haotianye@pku.edu.cn Chuanlong Xie Huawei Noah s Ark Lab xie.chuanlong@huawei.com Tianle Cai Peking University caitianle1998@pku.edu.cn Ruichen Li Peking University xk-lrc@pku.edu.cn Zhenguo Li Huawei Noah s Ark Lab Li.Zhenguo@huawei.com Liwei Wang Key Laboratory of Machine Perception, MOE, School of EECS, Institute for Artificial Intelligence, Peking University wanglw@cis.pku.edu.cn
Pseudocode Yes Algorithm 1: Model Selection
Open Source Code No Our experiments is conducted in Domain Bed: https://github.com/facebookresearch/Domain Bed. This statement indicates the use of an existing framework for conducting experiments, not an explicit release of the authors' own implementation code for their proposed methodology.
Open Datasets Yes We train our model on three benchmark OOD datasets (PACS [34], Office Home [59], VLCS [57])
Dataset Splits Yes Algorithm 1: Model Selection Input: available dataset Xavail = (Xtrain, Xval)... Accf compute validation accuracy of f using Xval
Hardware Specification No The paper does not specify any particular hardware details such as GPU models, CPU types, or memory specifications used for running the experiments.
Software Dependencies No The paper mentions 'Domain Bed' and 'GPU kernel density estimation' but does not provide specific version numbers for any software, libraries, or frameworks used in the experiments.
Experiment Setup Yes We train 200 models using different algorithms, penalties, learning rates, and epoch. For more details about the experiments, see Appendix 4.