Column Sampling Based Discrete Supervised Hashing

Authors: Wang-Cheng Kang, Wu-Jun Li, Zhi-Hua Zhou

AAAI 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Empirical results on datasets with semantic labels illustrate that COSDISH can outperform the state-of-the-art methods in real applications, such as image retrieval. Experiment We use real datasets to evaluate the effectiveness of our method.
Researcher Affiliation Academia Wang-Cheng Kang, Wu-Jun Li and Zhi-Hua Zhou National Key Laboratory for Novel Software Technology Collaborative Innovation Center of Novel Software Technology and Industrialization, Nanjing 210023 Department of Computer Science and Technology, Nanjing University, China kwc.oliver@gmail.com, {liwujun,zhouzh}@nju.edu.cn
Pseudocode Yes Algorithm 1 Discrete optimization in COSDISH
Open Source Code No The paper mentions supplementary material with a URL (http://cs.nju.edu.cn/lwj/paper/COSDISH_sup.pdf) but does not explicitly state that source code for the methodology is provided at this link or elsewhere.
Open Datasets Yes Two image datasets with semantic labels are used to evaluate our method and the other baselines. They are CIFAR-10 (Krizhevsky 2009) and NUS-WIDE (Chua et al. 2009).
Dataset Splits Yes As in LFH (Zhang et al. 2014), for all datasets we randomly choose 1000 points as validation set and 1000 points as query (test) set, with the rest of the points as training set.
Hardware Specification Yes All the experiments are conducted on a workstation with 6 Intel Xeon CPU cores and 48GB RAM.
Software Dependencies No The paper mentions using LBFGSB but does not provide specific version numbers for any software dependencies.
Experiment Setup Yes Unless otherwise stated, we set Tsto = 10, Talt = 3 and |Ω| = q in our experiments. Furthermore, in our experiments we find that our algorithm is not sensitive to the initialization. Hence, we adopt random initialization in this paper.