Timezone: »
Curriculum learning is a technique to improve a model performance and generalization based on the idea that easy samples should be presented before difficult ones during training. While it is generally complex to estimate a priori the difficulty of a given sample, recent works have shown that curriculum learning can be formulated dynamically in a self-supervised manner. The key idea is to somehow estimate the importance (or weight) of each sample directly during training based on the observation that easy and hard samples behave differently and can therefore be separated. However, these approaches are usually limited to a specific task (e.g., classification) and require extra data annotations, layers or parameters as well as a dedicated training procedure. We propose instead a simple and generic method that can be applied to a variety of losses and tasks without any change in the learning procedure. It consists in appending a novel loss function on top of any existing task loss, hence its name: the SuperLoss. Its main effect is to automatically downweight the contribution of samples with a large loss, i.e. hard samples, effectively mimicking the core principle of curriculum learning. As a side effect, we show that our loss prevents the memorization of noisy samples, making it possible to train from noisy data even with non-robust loss functions. Experimental results on image classification, regression, object detection and image retrieval demonstrate consistent gain, particularly in the presence of noise.
Author Information
Thibault Castells (Naver Labs)
Philippe Weinzaepfel (NAVER LABS Europe)
Jerome Revaud (Naver Labs Europe)
More from the Same Authors
-
2022 Poster: CroCo: Self-Supervised Pre-training for 3D Vision Tasks by Cross-View Completion »
Philippe Weinzaepfel · Vincent Leroy · Thomas Lucas · Romain BRÉGIER · Yohann Cabon · Vaibhav ARORA · Leonid Antsfeld · Boris Chidlovskii · Gabriela Csurka · Jerome Revaud -
2020 Poster: Hard Negative Mixing for Contrastive Learning »
Yannis Kalantidis · Mert Bulent Sariyildiz · Noe Pion · Philippe Weinzaepfel · Diane Larlus -
2019 Poster: R2D2: Reliable and Repeatable Detector and Descriptor »
Jerome Revaud · Cesar De Souza · Martin Humenberger · Philippe Weinzaepfel -
2019 Oral: R2D2: Reliable and Repeatable Detector and Descriptor »
Jerome Revaud · Cesar De Souza · Martin Humenberger · Philippe Weinzaepfel