Domain-Constraint Transfer Coding for Imbalanced Unsupervised Domain Adaptation

Authors: Yao-Hung Hubert Tsai, Cheng-An Hou, Wei-Yu Chen, Yi-Ren Yeh, Yu-Chiang Frank Wang

AAAI 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We perform comprehensive experiments on cross-domain classification with three different settings: standard UDA with balanced cross-domain data, UDA with mixed-domain data, and UDA with imbalanced cross-domain label numbers. Our experimental results would verify the effectiveness of our Dc TC in dealing with different cross-domain classification tasks.
Researcher Affiliation Academia 1Research Center for IT Innovation, Academia Sinica, Taipei, Taiwan 2The Robotics Institute, Carnegie Mellon University, Pittsburgh, PA 15213, USA 3Graduate Institute of Networking & Multimedia, National Taiwan University, Taipei, Taiwan 4Department of Mathematics, National Kaohsiung Normal University, Kaohsiung, Taiwan
Pseudocode Yes Algorithm 1 Domain-constraint Transfer Coding (Dc TC) for Imbalanced Unsupervised Domain Adaptation
Open Source Code No The paper does not provide any statement or link indicating that the source code for the described methodology is publicly available.
Open Datasets Yes We consider Office (Saenko et al. 2010) and Caltech256 (Griffin, Holub, and Perona 2007) datasets.
Dataset Splits No The paper mentions 'labeled training and unlabeled test data collected from source and target domains' but does not provide specific 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., CPU, GPU models, or memory) used for running the experiments.
Software Dependencies No The paper mentions applying 'De CAF6 features (Donahue et al. 2013)' but does not provide specific software library names with version numbers (e.g., Python 3.x, PyTorch 1.x) used for the implementation.
Experiment Setup Yes As for parameter selection, we set α = λ = 1, and feature dimension dc = 100.