Divide and Contrast: Source-free Domain Adaptation via Adaptive Contrastive Learning

Authors: Ziyi Zhang, Weikai Chen, Hui Cheng, Zhen Li, Siyuan Li, Liang Lin, Guanbin Li

NeurIPS 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments on Vis DA, Office-Home, and the more challenging Domain Net have verified the superior performance of Da C over current state-of-the-art approaches.
Researcher Affiliation Collaboration Ziyi Zhang1, Weikai Chen3, Hui Cheng2, Zhen Li4,5, Siyuan Li6, Liang Lin2, Guanbin Li2 1National Key Laboratory of Novel Software Technology, Nanjing University, Nanjing, China 2School of Computer Science and Engineering, Sun Yat-sen University, Guangzhou, China 3 Tencent America, 4 The Chinese University of Hong Kong, Shenzhen, China 5 Shenzhen Research Institute of Big Data, Shenzhen, China 6 AI Lab, School of Engineering, Westlake University, Hangzhou, China
Pseudocode Yes The overall algorithm of Da C is summarized in Appendix C.
Open Source Code Yes The code is available at https://github.com/Zye Zhang/Da C.git.
Open Datasets Yes We conduct experiments on three benchmark datasets: Office-Home [31], Vis DA-2017 [32], Domain Net [33].
Dataset Splits No The paper mentions using benchmark datasets but does not explicitly provide details on how the training, validation, and test splits were performed for their experiments.
Hardware Specification No The paper does not provide specific hardware details such as GPU/CPU models or memory used for running its experiments.
Software Dependencies No The paper does not provide specific software dependencies with version numbers (e.g., Python, PyTorch, or other libraries with versions).
Experiment Setup Yes The learning rate for the backbone is set as 2e-2 on Office-Home, 5e-4 on Vis DA, and 1e-2 on Domain Net. We train 30 epochs for Office-Home, 60 epochs for Vis DA, and 30 epochs for Domain Net.