Information-Theoretic Domain Adaptation Under Severe Noise Conditions

Authors: Wei Wang, Hao Wang, Zhi-Yong Ran, Ran He

AAAI 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments on two real-world domain adaptation tasks demonstrate the superiority of our method. and In this section, we evaluate the proposed method in two domain adaptation related applications: 1) object recognition and 2) face recognition.
Researcher Affiliation Collaboration Wei Wang,1 Hao Wang,2 Zhi-Yong Ran,3 Ran He4 1Institute of Software, Chinese Academy of Sciences, Beijing 100190, China. 2360 Search Lab, Qihoo 360, Beijing 100190, China. 3Chongqing University of Posts and Telecommunications, Chongqing 400065, China. 4Institute of Automation, Chinese Academy of Sciences, Beijing 100190, China.
Pseudocode Yes Algorithm 1 Robust Information-Theoretic Domain Adaptation (RIDA)
Open Source Code No The paper does not provide an explicit statement about releasing the source code for the described methodology, nor does it include a direct link to a code repository.
Open Datasets Yes COIL-20 (Nene et al. 1996) dataset contains 1,440 images from 20 objects. and CMU-PIE (Sim, Baker, and Bsat 2002) is a benchmark face dataset which includes 41,368 face images from 68 individuals with different poses, illuminations and facial expressions.
Dataset Splits No The paper describes partitioning datasets into source and target domains (e.g., COIL1 and COIL2 for COIL-20) and constructing cross-domain datasets, but it does not specify explicit train/validation/test splits, percentages, or sample counts needed for reproduction. It refers to 'labeled source data' and 'unlabeled target data' but not clear dataset split ratios.
Hardware Specification No The paper does not provide specific hardware details such as GPU/CPU models, processor types, or memory amounts used for running the experiments. It only discusses the general experimental setup without hardware specifications.
Software Dependencies No The paper does not provide specific software dependency details, such as library names with version numbers, that would be needed to replicate the experiment.
Experiment Setup Yes Our method involves three parameters: λ1, λ2 and σ. Across the experiments, we set these parameters by searching the values in the range [10 3, 103]. In general, our method is found to be robust to these parameters.