Cross-Modal Image Clustering via Canonical Correlation Analysis

Authors: Cheng Jin, Wenhui Mao, Ruiqi Zhang, Yuejie Zhang, Xiangyang Xue

AAAI 2015 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Very positive results were obtained in our experiments using a large quantity of public data.
Researcher Affiliation Academia Cheng Jin, Wenhui Mao, Ruiqi Zhang, Yuejie Zhang, Xiangyang Xue School of Computer Science, Shanghai Key Laboratory of Intelligent Information Processing, Fudan University, Shanghai, China {jc, 13210240099, 12210240075, yjzhang, xyxue}@fudan.edu.cn
Pseudocode Yes Algorithm 1 AHKMC(multimodal feature point set MFS) [...], Algorithm 2 AKMC(multi-modal feature point set MFS, initial center set C) [...], Algorithm 3 UPDATE(Center(i), C) [...]
Open Source Code No The paper does not provide any explicit statement about releasing open-source code or a link to a code repository.
Open Datasets Yes Our dataset is established based on two benchmark datasets of Corel30k and Nus-Wide.
Dataset Splits No The paper mentions using Corel30k and Nus-Wide datasets for experiments but does not provide specific details about training, validation, or test splits (e.g., percentages, sample counts, or explicit standard split names). It only discusses the effect of parameter alpha and feature dimensions on performance.
Hardware Specification No The paper does not provide any specific hardware details such as CPU, GPU models, or memory specifications used for running the experiments. It only implies the need for handling 'large-scale annotated image collections' and 'huge amounts of data'.
Software Dependencies No The paper discusses various techniques and algorithms (e.g., SIFT, BoF, LLC, CCA, k-means) but does not provide specific version numbers for any software dependencies or libraries used in the implementation.
Experiment Setup Yes To show the effect of the important parameter α in Formula (13) on the whole clustering performance, we compare the performance rising speeds for different settings of α, as shown in Figure 3. Thus we set α as 0.3 to make a trade-off for better performance. Compared the results for different numbers of feature dimension, we can observe that with the dimension increasing, the performance gradually becomes pretty good.