Cross-Domain Metric Learning Based on Information Theory
Authors: Hao Wang, Wei Wang, Chen Zhang, Fanjiang Xu
AAAI 2014 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experiments in two real-world applications demonstrate the effectiveness of our proposed method. In this section, we evaluate the proposed method in two metric learning related applications: 1) face recognition and 2) text classification. |
| Researcher Affiliation | Academia | 1. State Key Laboratory of Computer Science 2. Science and Technology on Integrated Information System Laboratory Institute of Software, Chinese Academy of Sciences, Beijing 100190, China 3. Department of Automation, University of Science and Technology of China |
| Pseudocode | No | The paper describes the proposed algorithm in detail but does not provide structured pseudocode or an explicitly labeled algorithm block. |
| Open Source Code | No | The paper does not provide any concrete access to source code for the methodology described, nor does it state that the code is released or available in supplementary materials. |
| Open Datasets | Yes | FERET (Phillips et al. 2000) and YALE (Belhumeur, Hespanha, and Kriegman 1997) are two public face data sets. 20-Newsgroups and Reuters-21578 are two benchmark text data sets widely used for evaluating the transfer learning algorithms (Dai et al. 2007b; Li, Jin, and Long 2012; Pan et al. 2011). The preprocessed version of Reuters-21578 on the web site (http://www.cse.ust.hk/TL/index.html) is used which contains three cross-domain data sets: orgs vs people, orgs vs place and people vs place. |
| Dataset Splits | No | The paper describes the construction of source and target domains for the experiments (e.g., 'the source domain set is YALE, and the target domain set consists of 100 individuals randomly selected from FERET'). However, it does not provide specific training/validation/test dataset splits (e.g., percentages, sample counts, or explicit mention of a validation set) needed to reproduce the partitioning of the data beyond source/target definitions. |
| Hardware Specification | Yes | The experiments are carried out on a single machine with Intel Core 2 Quad @ 2.40Ghz and 10 GB of RAM running 64-bit Windows 7. |
| Software Dependencies | No | The paper mentions standard classifiers like KNN and SVM but does not provide specific version numbers for any software components, libraries, or solvers used in the experiments. |
| Experiment Setup | Yes | CDML involves four parameters: σd, σ, µ and k. Specifically, we set σd by searching the values among {0.1, 1, 10}, σ among {0.1, 1, 10} and µ among {0.01, 0.1, 1, 10}. The neighborhood size k for CDML is 3. |