MADG: Margin-based Adversarial Learning for Domain Generalization

Authors: Aveen Dayal, Vimal K B, Linga Reddy Cenkeramaddi, C Mohan, Abhinav Kumar, Vineeth N Balasubramanian

NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We extensively experiment with the MADG model on popular real-world DG datasets, VLCS, PACS, Office Home, Domain Net, and Terra Incognita. We evaluate the proposed algorithm on Domain Bed s benchmark and observe consistent performance across all the datasets.
Researcher Affiliation Academia Aveen Dayal Indian Institute of Technology Hyderabad ai21resch11003@iith.ac.in Vimal K B Indian Institute of Technology Hyderabad vimalkb96@gmail.com Linga Reddy Cenkeramaddi University of Agder linga.cenkeramaddi@uia.no C Krishna Mohan Indian Institute of Technology Hyderabad ckm@cse.iith.ac.in Abhinav Kumar Indian Institute of Technology Hyderabad abhinavkumar@ee.iith.ac.in Vineeth N Balasubramanian Indian Institute of Technology Hyderabad vineethnb@cse.iith.ac.in
Pseudocode Yes Algorithm 1 Margin-based adversarial learning for Domain Generalization (MADG)
Open Source Code No No explicit statement or link providing concrete access to the source code for the methodology.
Open Datasets Yes We perform an extensive evaluation on five benchmark DG datasets for image classification: VLCS [30], PACS [31], Office Home (OH) [32], Terra Incognita (TI) [33] and Domain Net (DN) [34].
Dataset Splits Yes We follow [53] in using Test-domain validation procedure for hyperparameter selection.
Hardware Specification No The paper mentions 'GPU(GB)' in Table 2 for computational cost, but does not provide specific GPU models (e.g., NVIDIA A100), CPU models, or other detailed hardware specifications used for experiments.
Software Dependencies No The paper mentions using Resnet50 architecture and stochastic gradient descent, but does not provide specific software versions for libraries, frameworks, or languages (e.g., PyTorch version, Python version).
Experiment Setup Yes All other implementation details including hyperparameters such as learning rate, margin, and weight decay are provided in the Appendix.