MentorNet: Learning Data-Driven Curriculum for Very Deep Neural Networks on Corrupted Labels

TL;DR

MentorNet is a neural network that learns a data-driven curriculum to guide deep networks in training on corrupted labels, significantly improving generalization and achieving state-of-the-art results on noisy large-scale datasets.

Contribution

We introduce MentorNet, a neural network that dynamically learns sample weighting schemes to improve training on corrupted labels, surpassing existing curriculum methods.

Findings

Significantly improved generalization on corrupted datasets.

Achieved state-of-the-art results on WebVision benchmark.

Demonstrated effectiveness of data-driven curriculum learning.

Abstract

Recent deep networks are capable of memorizing the entire data even when the labels are completely random. To overcome the overfitting on corrupted labels, we propose a novel technique of learning another neural network, called MentorNet, to supervise the training of the base deep networks, namely, StudentNet. During training, MentorNet provides a curriculum (sample weighting scheme) for StudentNet to focus on the sample the label of which is probably correct. Unlike the existing curriculum that is usually predefined by human experts, MentorNet learns a data-driven curriculum dynamically with StudentNet. Experimental results demonstrate that our approach can significantly improve the generalization performance of deep networks trained on corrupted training data. Notably, to the best of our knowledge, we achieve the best-published result on WebVision, a large benchmark containing 2.2 million images of real-world noisy labels. The code are at https://github.com/google/mentornet