Parameter Free Large Margin Nearest Neighbor for Distance Metric Learning
Authors: Kun Song, Feiping Nie, Junwei Han, Xuelong Li
AAAI 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments on different data sets with various sizes and difficulties are conducted, and the results have shown that, compared with LMNN, PFLMNN achieves better classification results. |
| Researcher Affiliation | Academia | School of Automation, Northwestern Polytechnical University, Xi an, 710072, Shaanxi, P. R. China School of Computer Science and Center for OPTIMAL, Northwestern Polytechnical University, Xi an,710072, P. R. China Center for OPTIMAL, State Key Laboratory of Transient Optics and Photonics, Xi an Institute of Optics and Precision Mechanics, Chinese Academy of Sciences, Xi an 710119, Shaanxi, P. R. China. |
| Pseudocode | Yes | Algorithm 1 PFLMNN |
| Open Source Code | No | The paper does not provide any concrete access to source code for the methodology described, such as a repository link or an explicit code release statement. |
| Open Datasets | Yes | All the experiments are conducted on six data sets with different sizes and difficulties, i.e. Iris, Wine, Isolet, AT&T1, Coil100 (Nene, Nayar, and Hiroshi 1996) and USPS (Hull 1994). Among those data sets, the Wine, Iris and Isolet are taken from the UCI Machine learning Repository2. AT&T, Coil100, USPS are the hunman face image data sets, objective image data set and the hand-writing digit data, respectively. |
| Dataset Splits | Yes | All of the experiment results are averaged over several runs of randomly generated 70/30 splits of the data. Each experiment runs 50 times independently. For the purpose of cross-validation, the training sets would be partitioned into training and validation sets at 80/20 |
| Hardware Specification | No | The paper does not provide specific details about the hardware used for running the experiments, such as CPU or GPU models. |
| Software Dependencies | No | The paper does not provide specific software dependency details, such as library names with version numbers, needed to replicate the experiments. |
| Experiment Setup | Yes | In all of the experiments reported here, the parameter λ of LMNN is tuned by 5-fold cross validation (For the purpose of cross-validation, the training sets would be partitioned into training and validation sets at 80/20), and the searching grid is set at {0.02, 0.04, , 1}. The nearest neighbors number, i.e. k is set by cross-validation as recommended in (Weinberger and Saul 2009) for all the methods in our experiment. For KPFLMNN, KPCA, KRVML, the Gaussian RBF kernel is adopted, and the variance of the RBF kernel is set as the mean of Euclidean distances between all pairwise samples in the training set. |