Clarinet: A One-step Approach Towards Budget-friendly Unsupervised Domain Adaptation

Authors: Yiyang Zhang, Feng Liu, Zhen Fang, Bo Yuan, Guangquan Zhang, Jie Lu

IJCAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments show that CLARINET significantly outperforms a series of competent baselines.
Researcher Affiliation Academia 1Shenzhen International Graduate School, Tsinghua University 2Centre for Artificial Intelligence, University of Technology Sydney
Pseudocode Yes Algorithm 1 CLARINET: One-step BFUDA Approach
Open Source Code Yes The code of CLARINET is available at github.com/Yiyang98/BFUDA.
Open Datasets Yes Based on five commonly used datasets: MNIST (M), USPS (U), SVHN (S), MNIST-M (m) and SYN-DIGITS (Y), we verify efficacy of CLARINET on 6 BFUDA tasks
Dataset Splits No The paper states it follows "standard protocols for unsupervised domain adaptation" but does not explicitly provide percentages or counts for training, validation, or test dataset splits.
Hardware Specification No The paper does not provide specific details about the hardware (e.g., GPU/CPU models, memory) used for running the experiments.
Software Dependencies No The paper mentions implementing methods "by PyTorch" but does not specify a version number for PyTorch or any other software dependencies.
Experiment Setup Yes The batch size is set to 128 and the number of epochs is set to 500. SGD optimizer (momentum = 0.9, weight decay = 5 10 5) is with an initial learning rate of 0.005 in adversarial network and 5 10 5 in classifier. In mapping function T, l is set to 0.5.