Customized Multiple Clustering via Multi-Modal Subspace Proxy Learning

Authors: Jiawei Yao, Qi Qian, Juhua Hu

NeurIPS 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Our method consistently outperforms existing baselines across a broad set of datasets in visual multiple clustering tasks.
Researcher Affiliation Collaboration Jiawei Yao1 Qi Qian2 Juhua Hu1 1 School of Engineering and Technology, University of Washington, Tacoma, WA 98402, USA 2 Zoom Video Communications
Pseudocode No The paper does not include a clearly labeled pseudocode or algorithm block.
Open Source Code Yes Our code is available at https://github.com/Alexander-Yao/Multi-Sub.
Open Datasets Yes we evaluate the proposed method on almost all publicly available visual datasets commonly used in multiple clustering tasks Yu et al. [2024], including Stanford Cars Yao et al. [2024], Card Yao et al. [2023], CMUface Günnemann et al. [2014], Flowers Yao et al. [2024], Fruit Hu et al. [2017] and Fruit360 Yao et al. [2023]. Additionally, we created a multiple clustering dataset from CIFAR-10 Krizhevsky et al. [2009]
Dataset Splits No The paper mentions training for 1000 epochs and tuning hyperparameters, but does not explicitly provide training/validation/test dataset splits with percentages or sample counts.
Hardware Specification Yes The experiments are performed on four NVIDIA Ge Force RTX 2080 Ti GPUs.
Software Dependencies No The paper mentions using CLIP and GPT-4, and the Adam optimizer, but does not provide specific version numbers for any software libraries or dependencies used in the experiments.
Experiment Setup Yes For each user s preference, we train the model for 1000 epochs using Adam optimizer with a momentum of 0.9. We tune all the hyper-parameters based on the loss score of Multi-Sub, where the learning rate is selected from {1e-1,5e-2,1e-2,5e-3,1e-3,5e-4}, weight decay is chosen from {5e-4,1e-4,5e-5,1e-5, 0} for all the experiments.