Attentive Recurrent Comparators

Abstract

Rapid learning requires flexible representations to quickly adopt to new evidence. We develop a novel class of models called Attentive Recurrent Comparators (ARCs) that form representations of objects by cycling through them and making observations. Using the representations extracted by ARCs, we develop a way of approximating a \textit{dynamic representation space} and use it for one-shot learning. In the task of one-shot classification on the Omniglot dataset, we achieve the state of the art performance with an error rate of 1.5\%. This represents the first super-human result achieved for this task with a generic model that uses only pixel information.

References in corpus (4)

• Deep Learning in Neural Networks: An Overview
• Recurrent Models of Visual Attention
• DRAW: A Recurrent Neural Network For Image Generation
• Multiple Object Recognition with Visual Attention

Cited by in corpus (29)

• Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks
• Generalizing from a Few Examples: A Survey on Few-Shot Learning
• TADAM: Task dependent adaptive metric for improved few-shot learning
• A Two-Stage Approach to Few-Shot Learning for Image Recognition
• Sample Efficient Adaptive Text-to-Speech
• Generative Adversarial Residual Pairwise Networks for One Shot Learning
• Few-Shot Learning Through an Information Retrieval Lens
• Few-shot Autoregressive Density Estimation: Towards Learning to Learn Distributions
• DeepEMD: Differentiable Earth Mover's Distance for Few-Shot Learning
• Auto-Meta: Automated Gradient Based Meta Learner Search
• How to Train Your MAML to Excel in Few-Shot Classification
• Image compression optimized for 3D reconstruction by utilizing deep neural networks
• Toward Multimodal Model-Agnostic Meta-Learning
• Similarity R-C3D for Few-shot Temporal Activity Detection
• Revisiting Few-shot Activity Detection with Class Similarity Control
• Deep Co-attention based Comparators For Relative Representation Learning in Person Re-identification
• Evolutionary Architecture Search For Deep Multitask Networks
• Learning a Prior over Intent via Meta-Inverse Reinforcement Learning
• Zero-Shot Learning with Knowledge Enhanced Visual Semantic Embeddings
• Learning Task-General Representations with Generative Neuro-Symbolic Modeling
• Cross-Modulation Networks for Few-Shot Learning
• Semi-Supervised and Active Few-Shot Learning with Prototypical Networks
• One-Shot Segmentation in Clutter
• A Relational Model for One-Shot Classification
• Open-Ended Content-Style Recombination Via Leakage Filtering
• COFGA: Classification Of Fine-Grained Features In Aerial Images
• Meta R-CNN : Towards General Solver for Instance-level Few-shot Learning
• Image Matching via Loopy RNN
• Automatic detection of rare pathologies in fundus photographs using few-shot learning