论文信息 - A Deep Generative Model for Semi-Supervised Classification with Noisy Labels

A Deep Generative Model for Semi-Supervised Classification with Noisy Labels

Class labels are often imperfectly observed, due to mistakes and to genuine ambiguity among classes. We propose a new semi-supervised deep generative model that explicitly models noisy labels, called the Mislabeled VAE (M-VAE). The M-VAE can perform better than existing deep generative models which do not account for label noise. Additionally, the derivation of M-VAE gives new theoretical insights into the popular M1+M2 semi-supervised model.

[1] Xiaogang Wang,et al. Learning from massive noisy labeled data for image classification , 2015, 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[2] Nir Yosef,et al. Bayesian Inference for a Generative Model of Transcriptome Profiles from Single-cell RNA Sequencing , 2018, bioRxiv.

[3] José Miguel Hernández-Lobato,et al. Bayesian Semisupervised Learning with Deep Generative Models , 2017, 1706.09751.

[4] Arash Vahdat,et al. Toward Robustness against Label Noise in Training Deep Discriminative Neural Networks , 2017, NIPS.

[5] S. Linnarsson,et al. Cell types in the mouse cortex and hippocampus revealed by single-cell RNA-seq , 2015, Science.

[6] Max Welling,et al. Semi-supervised Learning with Deep Generative Models , 2014, NIPS.