Quantifying and Improving Transferability in Domain Generalization

Authors: Guojun Zhang, Han Zhao, Yaoliang Yu, Pascal Poupart

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

Reproducibility Variable Result LLM Response
Research Type Experimental Empirically, we evaluate the transferability of the feature embeddings learned by existing algorithms for domain generalization. Experimental results show that the proposed algorithm achieves consistent improvement over many state-of-the-art algorithms, corroborating our theoretical findings.
Researcher Affiliation Academia Guojun Zhang School of Computer Science University of Waterloo Vector Institute guojun.zhang@uwaterloo.ca Han Zhao Department of Computer Science University of Illinois at Urbana-Champaign hanzhao@illinois.edu Yaoliang Yu School of Computer Science University of Waterloo Vector Institute yaoliang.yu@uwaterloo.ca Pascal Poupart School of Computer Science University of Waterloo Vector Institute ppoupart@uwaterloo.ca
Pseudocode Yes Algorithm 1: Algorithm for evaluating transferability among multiple domains and Algorithm 2: Transfer algorithm for domain generalization
Open Source Code Yes Code available at https://github.com/Gordon-Guojun-Zhang/Transferability-NeurIPS2021.
Open Datasets Yes We test it over various benchmark datasets, including Rotated MNIST, PACS, Office-Home and WILDS-FMo W. [23] Pang Wei Koh, Shiori Sagawa, Henrik Marklund, Sang Michael Xie, Marvin Zhang, Akshay Balsubramani, Weihua Hu, Michihiro Yasunaga, Richard Lanas Phillips, Irena Gao, et al. WILDS: A benchmark of in-the-wild distribution shifts. ar Xiv preprint ar Xiv:2012.07421, 2020.
Dataset Splits No The paper mentions using standard benchmarks and refers to Appendix C for detailed experimental settings, but the main text does not provide specific train/validation/test dataset split percentages or sample counts for reproduction.
Hardware Specification No The paper mentions 'Vector Institute for providing the GPU cluster' but does not specify exact GPU models, CPU models, memory amounts, or other detailed computer specifications used for running experiments.
Software Dependencies No The paper does not explicitly list specific software dependencies with their version numbers (e.g., programming languages, libraries, frameworks) used in the experiments.
Experiment Setup Yes We take lr = 0.01 for Rotated MNIST and lr = 0.001 for other datasets. We used Adam optimizer with a learning rate of 10 3 for the feature extractor and 10 2 for the classifier. The batch size is 64. We train for 10 epochs. We use ReLU as activation function.