Automated Curriculum Learning for Neural Networks

Abstract

We introduce a method for automatically selecting the path, or syllabus, that a neural network follows through a curriculum so as to maximise learning efficiency. A measure of the amount that the network learns from each data sample is provided as a reward signal to a nonstationary multi-armed bandit algorithm, which then determines a stochastic syllabus. We consider a range of signals derived from two distinct indicators of learning progress: rate of increase in prediction accuracy, and rate of increase in network complexity. Experimental results for LSTM networks on three curricula demonstrate that our approach can significantly accelerate learning, in some cases halving the time required to attain a satisfactory performance level.

References in corpus (1)

• Weight Uncertainty in Neural Networks

Cited by in corpus (23)

• GradNorm: Gradient Normalization for Adaptive Loss Balancing in Deep Multitask Networks
• Hindsight Experience Replay
• Intrinsically Motivated Goal Exploration Processes with Automatic Curriculum Learning
• Curriculum Learning by Transfer Learning: Theory and Experiments with Deep Networks
• Automatic Goal Generation for Reinforcement Learning Agents
• Reverse Curriculum Generation for Reinforcement Learning
• Curriculum Learning and Minibatch Bucketing in Neural Machine Translation
• Illuminating Generalization in Deep Reinforcement Learning through Procedural Level Generation
• Unsupervised Meta-Learning for Reinforcement Learning
• Ray Interference: a Source of Plateaus in Deep Reinforcement Learning
• Unicorn: Continual Learning with a Universal, Off-policy Agent
• Adaptive Scheduling for Multi-Task Learning
• Energy-Based Hindsight Experience Prioritization
• Interactive Music Generation with Positional Constraints using Anticipation-RNNs
• Label-PEnet: Sequential Label Propagation and Enhancement Networks for Weakly Supervised Instance Segmentation
• Dynamic Multi-Level Multi-Task Learning for Sentence Simplification
• Adversarial Environment Generation for Learning to Navigate the Web
• AutoSeM: Automatic Task Selection and Mixing in Multi-Task Learning
• Multiple-objective Reinforcement Learning for Inverse Design and Identification
• Competitive Experience Replay
• Towards Scheduling Federated Deep Learning using Meta-Gradients for Inter-Hospital Learning
• An Empirical Comparison of Syllabuses for Curriculum Learning
• Integrating Feature and Image Pyramid: A Lung Nodule Detector Learned in Curriculum Fashion