A Survey on Curriculum Learning
arXiv:2010.13166
Abstract
Curriculum learning (CL) is a training strategy that trains a machine learning model from easier data to harder data, which imitates the meaningful learning order in human curricula. As an easy-to-use plug-in, the CL strategy has demonstrated its power in improving the generalization capacity and convergence rate of various models in a wide range of scenarios such as computer vision and natural language processing etc. In this survey article, we comprehensively review CL from various aspects including motivations, definitions, theories, and applications. We discuss works on curriculum learning within a general CL framework, elaborating on how to design a manually predefined curriculum or an automatic curriculum. In particular, we summarize existing CL designs based on the general framework of Difficulty Measurer+Training Scheduler and further categorize the methodologies for automatic CL into four groups, i.e., Self-paced Learning, Transfer Teacher, RL Teacher, and Other Automatic CL. We also analyze principles to select different CL designs that may benefit practical applications. Finally, we present our insights on the relationships connecting CL and other machine learning concepts including transfer learning, meta-learning, continual learning and active learning, etc., then point out challenges in CL as well as potential future research directions deserving further investigations.
arXiv comment20 pages, 7 figures, Accepted by IEEE Transactions on Pattern Analysis and Machine Intelligence 2021 (TPAMI)
References in corpus (21)
- SMOTE: Synthetic Minority Over-sampling Technique
- A continual learning survey: Defying forgetting in classification tasks
- Scheduled Sampling for Sequence Prediction with Recurrent Neural Networks
- Learning to Reweight Examples for Robust Deep Learning
- Training Deep Neural Networks on Noisy Labels with Bootstrapping
- Learning to Execute
- Online Batch Selection for Faster Training of Neural Networks
- Active Self-Paced Learning for Cost-Effective and Progressive Face Identification
- Curriculum Learning by Transfer Learning: Theory and Experiments with Deep Networks
- On The Power of Curriculum Learning in Training Deep Networks
- Curriculum Learning and Minibatch Bucketing in Neural Machine Translation
- An Empirical Exploration of Curriculum Learning for Neural Machine Translation
- Variance Reduction in SGD by Distributed Importance Sampling
- Learning to Teach with Dynamic Loss Functions
- Pseudo-Labeling Curriculum for Unsupervised Domain Adaptation
- Breaking the Curse of Space Explosion: Towards Efficient NAS with Curriculum Search
- Visualizing and Understanding Curriculum Learning for Long Short-Term Memory Networks
- ScreenerNet: Learning Self-Paced Curriculum for Deep Neural Networks
- Simple and Effective Curriculum Pointer-Generator Networks for Reading Comprehension over Long Narratives
- Self-Paced Multi-Task Learning
- Self-Paced Contextual Reinforcement Learning