Multidimensional Scaling on Multiple Input Distance Matrices

Authors: Song Bai, Xiang Bai, Longin Jan Latecki, Qi Tian

AAAI 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental results on synthetic as well as real datasets demonstrate the effectiveness of MVMDS.
Researcher Affiliation Academia Song Bai,1 Xiang Bai,1 Longin Jan Latecki,2 Qi Tian3 1Huazhong University of Science and Technology 2Temple University, 3University of Texas at San Antonio {songbai, xbai}@hust.edu.cn, latecki@temple.edu, qitian@cs.utsa.edu
Pseudocode Yes Algorithm 1: Multi-View Multidimensional Scaling.
Open Source Code No The paper does not include any explicit statement about providing open-source code for the methodology, nor does it provide a link to a code repository.
Open Datasets Yes Two image benchmark datasets, i.e., Microsoft Research Cambridge Volume 1 (MSRC-v1) (Winn and Jojic 2005), Caltech-101 dataset (Fei-Fei, Fergus, and Perona 2007), are selected for performance comparisons.
Dataset Splits No The paper describes the datasets used and categories selected (e.g., '30 images per category' for MSRC-v1, '7 classes and 20 classes forming Caltech101-7 and Caltech101-20'), but it does not provide specific details on how these datasets were partitioned into training, validation, and test sets (e.g., percentages, sample counts, or references to predefined splits).
Hardware Specification No The paper does not provide any specific hardware details (e.g., GPU/CPU models, processor types, memory amounts, or detailed computer specifications) used for running its experiments.
Software Dependencies No The paper mentions software tools and algorithms used (e.g., K-means, SIFT, HOG, LBP, HSV, GIST) but does not provide specific version numbers for any of them.
Experiment Setup Yes All the visual features are L2 normalized, then Euclidean distance is used to measure the dissimilarity between images. All the comparisons are done by using MDS or the proposed MVMDS to project images into P = 10 dimensional space. Since the weight controller γ needs to be determined empirically, we conduct an exhaustive search in the interval (1, 10] with step size 0.5 to find its optimal value. The desired number of clusters is set to be equal to the natural number of categories in each dataset.