Positive-Unlabeled Learning from Imbalanced Data

Authors: Guangxin Su, Weitong Chen, Miao Xu

IJCAI 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In this section, we compare the performance of our proposed Imbalancednn PU with state-of-the-art PU methods on imbalanced datasets. We will show that state-of-the-art PU methods, although have been shown to be effective on balanced PU data, fails to be superior on imbalanced PU data. [...] Dataset. In previous PU work, datasets such as CIFAR10 have been widely used [...] Evaluation. The same as [Kiryo et al., 2017], we set the number of epochs to be 200. [...] Results without meta-learning. The experimental results of methods without meta-learning on F1 score are shown in Figure 2.
Researcher Affiliation Academia Guangxin Su1 , Weitong Chen1 , Miao Xu1,2 1The University of Queensland, Brisbane QLD4072, Australia 2RIKEN AIP, Tokyo 103-0027, Japan guangxinsu6@gmail.com, {weitong.chen, miao.xu}@uq.edu.au
Pseudocode Yes Algorithm 1 Imbalancednn PU Input: Training data P and U Parameter: class prior π and π , MAX_E Output: classifier bg(x; θ)
Open Source Code No The paper states 'All the codes are implemented in Python 3 and Pytorch 1.7' but does not provide a specific link or explicit statement about the public availability of their source code for the proposed method.
Open Datasets Yes In our tasks, following existing works to test the scalability of our proposal, we also use the CIFAR10 data. [...] 1https://www.cs.toronto.edu/ kriz/cifar.html
Dataset Splits No In each dataset, there are 50,000 training data and 10,000 test data as provided by the original CIFAR10. To make the training data into a PU learning problem, we follow [Kiryo et al., 2017] to sample 1,000 positive instances and treat them as P; all the training data are used as U, i.e., nu = 50,000.
Hardware Specification No All the codes are implemented in Python 3 and Pytorch 1.7, and running on a GPU server with CUDA 11.1.
Software Dependencies Yes All the codes are implemented in Python 3 and Pytorch 1.7, and running on a GPU server with CUDA 11.1.
Experiment Setup Yes For our proposed Imbalancednn PU, we set π = 0.5 and π = 0.1. We use the same network structure as [Kiryo et al., 2017], i.e., a 13-layer CNN with Re LU and Adam as the optimizer. We tune the hyper-parameters step size and weight decay by a grid select from {10-10, 10-9, . . . , 100} for all methods based on neural networks. All the other hyperparameters in the network are set as default. The same as [Kiryo et al., 2017], we set the number of epochs to be 200.