Semi-Supervised AUC Optimization Without Guessing Labels of Unlabeled Data

Authors: Zheng Xie, Ming Li

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

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental results indicate that the proposed methods outperform the existing methods.
Researcher Affiliation Academia Zheng Xie, Ming Li National Key Laboratory for Novel Software Technology, Nanjing University Collaborative Innovation Center for Novel Software Technology and Industrialization Nanjing 210023, China Email: {xiez, lim}@lamda.nju.edu.cn
Pseudocode Yes Algorithm 1 SAMULT
Open Source Code No No explicit statement regarding open-source code availability or a direct link is provided.
Open Datasets Yes We evaluate our methods on 20 widely-used datasets, including 18 datasets from UCI repository (Lichman 2013), and the ijcnn1 and the madelon (Guyon et al. 2005).
Dataset Splits Yes The parameters are chosen by grid search through a 5-fold cross validation. For each experiment, 10% of the data in the training set is labeled for SAMULT and SAMULTP+U to train the classifiers, and SAMULTP+U uses only positive and unlabeled data. Roughly 20% data is held out as the test set.
Hardware Specification No No specific hardware details (e.g., GPU/CPU models, memory) are provided for the experimental setup.
Software Dependencies No No specific software dependencies with version numbers are provided in the paper.
Experiment Setup Yes The parameters are chosen by grid search through a 5-fold cross validation. The number of base learners in SAMPURA is fixed on 20.