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

By Chelsea Finn et al

Published on June 10, 2017

Read the original document by opening this link in a new tab.

Table of Contents

1. Introduction
2. Model-Agnostic Meta-Learning
3. Species of MAML
4. Related Work

Summary

The paper proposes an algorithm for meta-learning that is model-agnostic and applicable to various learning problems. The goal is to train a model on a variety of tasks so that it can adapt quickly to new tasks with minimal training data. The method focuses on optimizing the model's parameters for fast adaptation by training it to be easy to fine-tune. It demonstrates state-of-the-art performance on few-shot image classification and regression tasks. The algorithm is general and can be applied across different model types and problem settings, including reinforcement learning.