Robust Classification under Covariate Shift with Application to Active Learning

Authors: Anqi Liu

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

Reproducibility Variable Result LLM Response
Research Type Experimental The RBA classifier is applied to active learning by considering the problem as a covariate shift prediction task and adopting pessimism about all uncertain properties of the conditional label distribution (Liu, Reyzin, and Ziebart 2015). Theoretically, this aligns model uncertainty with prediction loss on remaining unlabeled data points, better justifying the use of the model s label estimates within active learning label solicitation strategies. Moreover, thanks to the cost sensitive method developed recently (Asif et al. 2015), the active learning framework using adversarial prediction is tractable in the 0-1 loss minimization under covariate shift. So it is also applied to active learning in a workshop paper (Liu et al. 2015), even though further effort to improve the performance is still required.
Researcher Affiliation Academia Anqi Liu Department of Computer Science University of Illinois at Chicago Chicago, Illinois 60607 aliu33@uic.edu
Pseudocode No The paper does not contain any pseudocode or clearly labeled algorithm blocks.
Open Source Code No The paper does not provide any explicit statements about releasing source code for the methodology described, nor does it include links to a code repository.
Open Datasets No The paper mentions 'training data' in a general sense but does not specify any particular dataset used for experiments or provide information about its public availability or access.
Dataset Splits No The paper does not provide specific details about training/validation/test dataset splits, such as percentages, sample counts, or references to predefined splits.
Hardware Specification No The paper does not provide any specific hardware details (e.g., GPU/CPU models, memory, cloud instances) used for running its experiments.
Software Dependencies No The paper refers to methods and prior work by citation (e.g., 'cost sensitive method developed recently (Asif et al. 2015)'), but it does not list any specific software dependencies with version numbers (e.g., libraries, frameworks, solvers).
Experiment Setup No The paper discusses the overall research framework and approach but does not provide specific experimental setup details, such as hyperparameter values, training configurations, or system-level settings.