Rethinking the Value of Labels for Improving Class-Imbalanced Learning

Authors: Yuzhe Yang, Zhi Xu

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

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments on large-scale imbalanced datasets verify our theoretically grounded strategies, showing superior performance over previous state-of-the-arts.
Researcher Affiliation Academia Yuzhe Yang EECS Massachusetts Institute of Technology yuzhe@mit.edu Zhi Xu EECS Massachusetts Institute of Technology zhixu@mit.edu
Pseudocode No The paper does not contain any explicitly labeled pseudocode or algorithm blocks.
Open Source Code Yes Code is available at https://github.com/Yyz Harry/imbalanced-semi-self.
Open Datasets Yes We conduct thorough experiments on artificially created long-tailed versions of CIFAR-10 [7] and SVHN [36]... 80 Million Tiny Images [48] for CIFAR-10, and SVHN s own extra set [36]... Image Net-LT [33] and real-world dataset i Naturalist 2018 [24].
Dataset Splits No The paper mentions using standard benchmark datasets and states 'We follow [7,25,33] to evaluate models on corresponding balanced test datasets', implying predefined splits, but does not explicitly provide specific percentages, sample counts, or a detailed splitting methodology for training, validation, and test sets within the paper itself.
Hardware Specification No The paper does not provide specific hardware details such as GPU or CPU models used for running the experiments.
Software Dependencies No The paper does not provide specific software dependencies or version numbers for the libraries or frameworks used in the experiments.
Experiment Setup Yes In the classifier learning stage, we follow [7,25] to train all models for 200 epochs on CIFAR-LT, and 90 epochs on Image Net-LT and i Naturalist. We use Rotation [16] as SSP method on CIFAR-LT, and Mo Co [19] on Image Net-LT and i Naturalist.