Prototypical Networks for Few-shot Learning (2017-03-15T00:00:00.000000Z)

TL;DR

This work proposes Prototypical Networks for few-shot classification, and provides an analysis showing that some simple design decisions can yield substantial improvements over recent approaches involving complicated architectural choices and meta-learning.

Abstract

We propose Prototypical Networks for the problem of few-shot classification, where a classifier must generalize to new classes not seen in the training set, given only a small number of examples of each new class. Prototypical Networks learn a metric space in which classification can be performed by computing distances to prototype representations of each class. Compared to recent approaches for few-shot learning, they reflect a simpler inductive bias that is beneficial in this limited-data regime, and achieve excellent results. We provide an analysis showing that some simple design decisions can yield substantial improvements over recent approaches involving complicated architectural choices and meta-learning. We further extend Prototypical Networks to zero-shot learning and achieve state-of-the-art results on the CU-Birds dataset.

Authors

R. Zemel

11 papers

Kevin Swersky

5 papers

Jake Snell

2 papers

TL;DR

Abstract

Authors

References38 items

Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks

Optimization as a Model for Few-Shot Learning

Zero-Shot Recognition via Structured Prediction

A Discriminative Feature Learning Approach for Deep Face Recognition

Meta-Learning with Memory-Augmented Neural Networks

Matching Networks for One Shot Learning

Towards a Neural Statistician

Learning Deep Representations of Fine-Grained Visual Descriptions

Synthesized Classifiers for Zero-Shot Learning

Metric Learning with Adaptive Density Discrimination

Predicting Deep Zero-Shot Convolutional Neural Networks Using Textual Descriptions

Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift

Adam: A Method for Stochastic Optimization

Neural Turing Machines

Evaluation of output embeddings for fine-grained image classification

Going deeper with convolutions

Very Deep Convolutional Networks for Large-Scale Image Recognition

ImageNet Large Scale Visual Recognition Challenge

Stochastic Backpropagation and Approximate Inference in Deep Generative Models

Auto-Encoding Variational Bayes

Write a Classifier: Zero-Shot Learning Using Purely Textual Descriptions

Distance-Based Image Classification: Generalizing to New Classes at Near-Zero Cost

Metric Learning: A Survey

A Survey on Metric Learning for Feature Vectors and Structured Data

Label-Embedding for Attribute-Based Classification

ImageNet classification with deep convolutional neural networks

Caltech-UCSD Birds 200

A Deep Non-linear Feature Mapping for Large-Margin kNN Classification

Learning a Nonlinear Embedding by Preserving Class Neighbourhood Structure

Distance Metric Learning for Large Margin Nearest Neighbor Classification

Clustering with Bregman Divergences

Neighbourhood Components Analysis

Learning from one example through shared densities on transforms

Long Short-Term Memory

Learning Deep Parsimonious Representations

Siamese Neural Networks for One-Shot Image Recognition

One shot learning of simple visual concepts

Visualizing Data using t-SNE

Field of Study

Venue Information

Name

Type

URL

Alternate Names