Learning Algorithms for Active Learning

Abstract

We introduce a model that learns active learning algorithms via metalearning. For a distribution of related tasks, our model jointly learns: a data representation, an item selection heuristic, and a method for constructing prediction functions from labeled training sets. Our model uses the item selection heuristic to gather labeled training sets from which to construct prediction functions. Using the Omniglot and MovieLens datasets, we test our model in synthetic and practical settings.

Accepted for publication at ICML 2017

References in corpus (9)

• Adam: A Method for Stochastic Optimization
• Google's Neural Machine Translation System: Bridging the Gap between Human and Machine Translation
• High-Dimensional Continuous Control Using Generalized Advantage Estimation
• Matching Networks for One Shot Learning
• Weight Normalization: A Simple Reparameterization to Accelerate Training of Deep Neural Networks
• One-shot Learning with Memory-Augmented Neural Networks
• Gradient Estimation Using Stochastic Computation Graphs
• Reinforcement and Imitation Learning via Interactive No-Regret Learning
• Active One-shot Learning

Cited by in corpus (39)

• Generalizing from a Few Examples: A Survey on Few-Shot Learning
• Meta-Learning for Semi-Supervised Few-Shot Classification
• A Survey on Active Learning and Human-in-the-Loop Deep Learning for Medical Image Analysis
• A Survey of Active Learning for Text Classification using Deep Neural Networks
• Geometry in Active Learning for Binary and Multi-class Image Segmentation
• Graph Policy Network for Transferable Active Learning on Graphs
• Reinforced active learning for image segmentation
• Metalearning with Hebbian Fast Weights
• Discovering General-Purpose Active Learning Strategies
• Using Error Decay Prediction to Overcome Practical Issues of Deep Active Learning for Named Entity Recognition
• How to Train Your MAML to Excel in Few-Shot Classification
• Why Not to Use Zero Imputation? Correcting Sparsity Bias in Training Neural Networks
• Revisiting Meta-Learning as Supervised Learning
• Deep Active Learning: Unified and Principled Method for Query and Training
• Meta-Learning for Contextual Bandit Exploration
• Learning to Sample the Most Useful Training Patches from Images
• Relaxed Softmax for learning from Positive and Unlabeled data
• MedSelect: Selective Labeling for Medical Image Classification Combining Meta-Learning with Deep Reinforcement Learning
• Maximizing Information Gain in Partially Observable Environments via Prediction Reward
• On Interactive Machine Learning and the Potential of Cognitive Feedback
• Crowd Sourcing based Active Learning Approach for Parking Sign Recognition
• Distribution Aware Active Learning
• Visual Curiosity: Learning to Ask Questions to Learn Visual Recognition
• Quality meets Diversity: A Model-Agnostic Framework for Computerized Adaptive Testing
• On the Robustness of Active Learning
• Mindful Active Learning
• Towards Understanding the Behaviors of Optimal Deep Active Learning Algorithms
• Augmented Memory Networks for Streaming-Based Active One-Shot Learning
• Dialog as a Vehicle for Lifelong Learning
• Not Your Grandfathers Test Set: Reducing Labeling Effort for Testing
• Learning to Actively Learn: A Robust Approach
• Meta-Active Learning for Node Response Prediction in Graphs
• Learning a Policy for Opportunistic Active Learning
• Leveraging Motion Priors in Videos for Improving Human Segmentation
• MetAL: Active Semi-Supervised Learning on Graphs via Meta Learning
• A Structured Perspective of Volumes on Active Learning
• ImitAL: Learning Active Learning Strategies from Synthetic Data
• On Data Efficiency of Meta-learning
• Batch Decorrelation for Active Metric Learning