Domain-Shared Group-Sparse Dictionary Learning for Unsupervised Domain Adaptation

Authors: Baoyao Yang, Andy Ma, Pong Yuen

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

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental results on cross-domain face and object recognition show that the proposed method outperforms eight state-of-the-art unsupervised domain adaptation algorithms. (Abstract)
Researcher Affiliation Academia 1Department of Computer Science, Hong Kong Baptist University, Hong Kong 2School of Data and Computer Science, Sun Yat-sen University, China
Pseudocode No Not found. The paper describes algorithms verbally and mathematically but does not include structured pseudocode or algorithm blocks.
Open Source Code No Not found. The paper does not explicitly state that source code is available or provide a link to a repository.
Open Datasets Yes CMU-PIE dataset (Sim, Baker, and Bsat 2002) is used for experiments of face recognition across blur and illumination variations.
Dataset Splits No Not found. The paper describes source and target domains and test performance but does not specify separate validation splits or how they are used for hyperparameter tuning.
Hardware Specification No Not found. The paper does not provide any specific hardware details such as GPU/CPU models or other computing specifications used for running experiments.
Software Dependencies No Not found. The paper mentions using 'De CAF6 features' but does not specify any software names with version numbers for implementation or dependencies.
Experiment Setup Yes For learning the domain-shared group-sparse dictionary, the sizes of each sub-dictionary are 5 and 10 for face and object recognition, respectively. We set the number of remaining bases as 2 for experiments. The sparsity coefficient λ is set as 0.1, which follows the experimental settings in TSC (Long et al. 2013a). Hyper-parameter experiments are done to analyze the parameter sensitivity of parameters η, δ, μ and β in the optimization function (Equation (6)). For learning target classifier, the parameter γ is set as 1 to balance each term in Equation (21). (Section 4.1)