Some Noisy-label Learning, Robust Learning, Semi-supervised Learning and Training tricks implementation.
MNIST and CIFAR10 dataset baseline without label noise or training tricks.
Referring to Active Learning. Using several models to inference, calculating each sample's weighted average loss
Title: O2U-Net: A Simple Noisy Label Detection Approach for Deep Neural Networks
Paper: https://openaccess.thecvf.com/content_ICCV_2019/papers/Huang_O2U-Net_A_Simple_Noisy_Label_Detection_Approach_for_Deep_Neural_ICCV_2019_paper.pdf
Title: Distilling the Knowledge in a Neural Network
Paper: https://arxiv.org/pdf/1503.02531.pdf
Title: Mean teachers are better role models:Weight-averaged consistency targets improve semi-supervised deep learning
results
Paper: https://arxiv.org/pdf/1703.01780.pdf
Title: Active Bias: Training More Accurate Neural Networks by Emphasizing High Variance Samples
Paper: https://arxiv.org/pdf/1704.07433.pdf
Title: Decoupling “when to update” from “how to update”
Paper: https://arxiv.org/pdf/1706.02613.pdf
Title: mixup: BEYOND EMPIRICAL RISK MINIMIZATION
Paper: https://arxiv.org/pdf/1710.09412.pdf
Title: MentorNet: Learning Data-Driven Curriculum for Very Deep Neural Networks on Corrupted Labels
Paper: https://arxiv.org/pdf/1712.05055.pdf
Title: Co-teaching: Robust Training of Deep Neural Networks with Extremely Noisy Labels
Paper: https://arxiv.org/pdf/1804.06872.pdf
Title: Generalized Cross Entropy Loss for Training Deep Neural Networks with Noisy Labels
Paper: https://arxiv.org/pdf/1805.07836v4.pdf
Title: AN EMPIRICAL STUDY OF EXAMPLE FORGETTING DURING DEEP NEURAL NETWORK LEARNING
Paper: https://arxiv.org/pdf/1812.05159.pdf
Title: Meta-Weight-Net: Learning an Explicit Mapping For Sample Weighting
Paper: https://arxiv.org/pdf/1902.07379.pdf
Title: MixMatch: A Holistic Approach to Semi-Supervised Learning
Paper: https://arxiv.org/pdf/1905.02249v1.pdf
Title: Symmetric Cross Entropy for Robust Learning with Noisy Labels
Paper: https://arxiv.org/pdf/1908.06112.pdf
Title: NLNL: Negative Learning for Noisy Labels
Paper: https://arxiv.org/pdf/1908.07387.pdf
Title: SELF: LEARNING TO FILTER NOISY LABELS WITH SELF-ENSEMBLING
Paper: https://arxiv.org/pdf/1910.01842.pdf
Title: FixMatch: Simplifying Semi-Supervised Learning with Consistency and Confidence
Paper: https://arxiv.org/ftp/arxiv/papers/2001/2001.07685.pdf
Title: DIVIDEMIX: LEARNING WITH NOISY LABELS AS SEMI-SUPERVISED LEARNING
Paper: https://arxiv.org/pdf/2002.07494v1.pdf
Title: Do We Need Zero Training Loss After Achieving Zero Training Error?
Paper: https://arxiv.org/pdf/2002.08709.pdf
Title Normalized Loss Functions for Deep Learning with Noisy Labels
Paper: https://arxiv.org/pdf/2006.13554v1.pdf
Train using label smoothing
Train using CE at early stage and using MAE after
Filtering noise image using several augmentation