Decoder Choice Network for Meta-Learning

This repo contains code accompaning the paper. It includes code for running the few-shot supervised learning domain experiments, including sinusoid regression, Omniglot classification, and MiniImagenet classification.

Dependencies

This code requires the following:

• python 3.*
• Pytorch 0.4.1+

Data

For the Omniglot and MiniImagenet data, see the usage instructions in pre_data/omniglot_resized.py and pre_data/mini_images.py respectively.

Sinusoid

To run the code, see the usage instructions at the top of Sine/sinusoid_e.py.

Omniglot

Generate few-shot learning task.

$ cd Omni
$ python omniglot_make_task.py

To run the code, see examples in Omni/omni_caps.py and Omni/omni_fmeta.py. Automated training and test fuzzycaps4

$ python omni_caps.py

Automated training and test fuzzymeta4

$ python omni_fmeta.py

MiniImagenet

Generate few-shot learning task.

$ cd Mini
$ python miniimagenet_make_task.py

To run the code, see examples in Mini/mini_auto_1shot.py and Mini/mini_auto_5shot.py. Automated training and test fuzzymeta on 1-shot task.

$ python mini_auto_1shot.py

Automated training and test fuzzymeta on 5-shot task.

$ python mini_auto_5shot.py