Rethinking Class-Prior Estimation for Positive-Unlabeled Learning

Authors: Yu Yao, Tongliang Liu, Bo Han, Mingming Gong, Gang Niu, Masashi Sugiyama, Dacheng Tao

ICLR 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We run experiments on 2 synthetic datasets and 9 real word datasets
Researcher Affiliation Collaboration 1The University of Sydney 2Hong Kong Baptist University 3The University of Melbourne 4RIKEN AIP 5The University of Tokyo 6JD Explore Academy, China
Pseudocode Yes Algorithm 1 Re CPE
Open Source Code Yes We have also included an anonymous source code in our supplementary material.
Open Datasets Yes The real-world datasets are downloaded from the UCL machine learning database. Multi-class datasets are used as binary datasets by either grouping or ignoring classes.
Dataset Splits Yes We sample the validation set with 20% of the training data size.
Hardware Specification No The paper mentions training a neural network but does not specify any hardware details such as GPU/CPU models or specific computing resources used for the experiments.
Software Dependencies No The paper does not provide specific software dependencies with version numbers (e.g., library names like PyTorch or TensorFlow with their versions).
Experiment Setup Yes For all experiments, we employ a neural network with 2 hidden layers. Each hidden layer contains 50 hidden units. The batch normalization (Ioffe & Szegedy, 2015) is also employed. The stochastic gradient descent optimizer is used with the batch size 50. The network is trained for 350 epochs with a learning rate 0.01 and momentum 0. The weight decay is set to 1e 5.