Online Metric Learning for Multi-Label Classification

Authors: Xiuwen Gong, Dong Yuan, Wei Bao4012-4019

AAAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Comprehensive experiments on a number of benchmark multi-label datasets validate our theoretical approach and illustrate that our proposed online metric learning (OML) algorithm outperforms state-of-the-art methods.
Researcher Affiliation Academia Xiuwen Gong, Dong Yuan, Wei Bao Faculty of Engineering, The University of Sydney {xiuwen.gong, dong.yuan, wei.bao}@sydney.edu.au
Pseudocode Yes Algorithm 1 Online Metric Learning for Multi-Label Classification
Open Source Code No The paper states 'The codes are provided by the respective authors.' in the context of baseline methods, and provides a link to a dataset source, but there is no explicit statement or link for the open-source code of their own proposed methodology.
Open Datasets Yes To evaluate the performance of our proposed online metric learning algorithm, we conduct experiments on four benchmark datasets: Corel5k, Enron, Medical and Emotions. The statistics of these datasets can be found in website1. 1http://mulan.sourceforge.net
Dataset Splits No The paper states 'Initially, we keep 20% of data for nearest neighbor searching.' but does not provide specific train/validation/test dataset splits needed to reproduce the experiment.
Hardware Specification Yes All experiments are conducted on a workstation with 3.20GHz Intel CPU and 16GB main memory, running the Windows 10 platform.
Software Dependencies No The paper does not provide specific software names with version numbers (e.g., library or solver names with version numbers) needed to replicate the experiment.
Experiment Setup Yes In our experiment, the matrix V1 is initialized as a normal distributed random matrix. Initially, we keep 20% of data for nearest neighbor searching. In our experiment, M is set to 100000 and m is set to 0.00001, while k is set to 10. Parameter λ in OLANSGD is chosen from among {10 6, 10 5, , 100} using five-fold cross validation. We use the default parameter for OSML-ELM.