Provably Consistent Partial-Label Learning

Authors: Lei Feng, Jiaqi Lv, Bo Han, Miao Xu, Gang Niu, Xin Geng, Bo An, Masashi Sugiyama

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

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments on benchmark and real-world datasets validate the effectiveness of the proposed generation model and two PLL methods.
Researcher Affiliation Academia 1School of Computer Science and Engineering, Nanyang Technological University, Singapore 2School of Computer Science and Engineering, Southeast University, Nanjing, China 3Department of Computer Science, Hong Kong Baptist University, China 4The University of Queensland, Australia 5Center for Advanced Intelligence Project, RIKEN, Japan 6The University of Tokyo, Japan
Pseudocode Yes Algorithm 1 RC Algorithm; Algorithm 2 CC Algorithm
Open Source Code No The paper does not provide explicit links or statements about the availability of open-source code for the described methodology.
Open Datasets Yes We collect four widely used benchmark datasets including MNIST [38], Kuzushiji MNIST [12], Fashion-MNIST [63], and CIFAR-10 [37], and five datasets from the UCI Machine Learning Repository [37]. In addition, we also use five widely used real-world partially labeled datasets, including Lost [13], Bird Song [6], MSRCv2 [41], Soccer Player [69], Yahoo! News [24].
Dataset Splits Yes Hyper-parameters are selected so as to maximize the accuracy on a validation set (10% of the training set) of partially labeled data.
Hardware Specification No The paper does not provide specific details about the hardware used for running the experiments (e.g., GPU/CPU models, memory, etc.).
Software Dependencies No The paper mentions "Py Torch [56]" and "Adam [36]" but does not specify their version numbers, nor does it list multiple key software components with versions.
Experiment Setup Yes Hyper-parameters are selected so as to maximize the accuracy on a validation set (10% of the training set) of partially labeled data. We implement them using Py Torch [56] and use the Adam [36] optimizer with the mini-batch size set to 256 and the number of epochs set to 250.