Unsupervised Generative Adversarial Cross-Modal Hashing

Authors: Jian Zhang, Yuxin Peng, Mingkuan Yuan

AAAI 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments compared with 6 state-of-the-art methods on 2 widely-used datasets verify the effectiveness of our proposed approach.
Researcher Affiliation Academia Jian Zhang, Yuxin Peng, Mingkuan Yuan Institute of Computer Science and Technology, Peking University Beijing 100871, China pengyuxin@pku.edu.cn
Pseudocode No The paper describes the model components and their operations using mathematical equations and textual descriptions, but does not provide structured pseudocode or an algorithm block.
Open Source Code No The paper states 'We implement the proposed UGACH by tensorflow1' and provides a link to the general TensorFlow website, but does not offer concrete access (e.g., a specific repository link or a clear statement of code release) for their UGACH implementation.
Open Datasets Yes In the experiments, we conduct cross-modal hashing on 2 widely-used datasets: NUS-WIDE (Chua et al. 2009) and MIRFLICKR (Huiskes and Lew 2008).
Dataset Splits No The paper defines a 'retrieval database' used as a training set and a 'query set' for testing, but does not explicitly mention a separate validation set for the proposed UGACH method. It mentions a training set for 'supervised methods' (baselines), not for UGACH itself.
Hardware Specification No The paper does not provide specific details about the hardware used to run the experiments, such as GPU models, CPU types, or memory specifications.
Software Dependencies No The paper only mentions 'tensorflow1' but does not specify a version number or any other software dependencies with their versions.
Experiment Setup Yes The dimension of common representation layer is set to be 4096, while the hashing layer s dimension is set to be the same as hash code length. Moreover, we train the proposed UGACH in a mini-batch way and set the batch size as 64 for discriminative and generative models. We train the proposed UGACH iteratively. After the discriminative model is trained in 1 epoch, the generative model respectively will be trained in 1 epoch. The learning rate of UGACH is decreased by a factor of 10 each two epochs, while it is initialized as 0.01.