Skip to content

This repository implements the paper, Model-Agnostic Meta-Leanring for Fast Adaptation of Deep Networks.

Notifications You must be signed in to change notification settings

hiwonjoon/maml-tensorflow

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

13 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

MAML

Intro

This repository implements the paper, Model-Agnostic Meta-Leanring for Fast Adaptation of Deep Networks.

Requirements

  • Tensorflow (v1.3 or higher)
  • better_exceptions, tqdm, Pillow, etc.

Training statisitics

  • Sinusoid

    Sinusoide Result Red line - ground truth, Red dots - given observation, Blue line - predicted line after 1 sgd step

  • Omniglot

    • Omniglot Testing (multiple descent steps)

    Ominglot Result

    • valid_acc_{0,1} means accuracy after 1 and 2 SGD steps. The valid_acc is the accuracy after the weights are trained with 3 SGD steps.

Training

Download datasets

Downlaod Omniglot dataset from the link. Only images_background.zip and images_evalueation.zip are required.

Unzip on the directory (repository)/datasets/omniglot/, so the directory shoud looks like (repo)/datasets/omniglot/{images_background,images_evaluation}.

Run train

  • Run sinusoide: python sinusoide.py
  • Run omniglot: python omniglot.py

Change the hyperparameters accordingly as you want. Please check at the bottom of each script.

TODO

  • Mini Imagenet Training
  • Robotic Simulation.

Acknowledgement

  • Author's original implementation: link

About

This repository implements the paper, Model-Agnostic Meta-Leanring for Fast Adaptation of Deep Networks.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages