This video walks through an implementation of Reptile in Keras using the Omniglot dataset. I was really inspired by this example, I think the Omniglot challenge of dynamically being able to recombine characters to form new alphabets is an incredibly interesting problem, connecting Human and Artificial Intelligence. I hope you found this example interesting as well, please check out the rest of the Keras Code Example playlist!
Content Links:
Few-shot learning with reptile: https://keras.io/examples/vision/rept...
On First-Order Meta Learning: https://arxiv.org/pdf/1803.02999.pdf
MAML: https://arxiv.org/pdf/1703.03400.pdf
Generative Teaching Networks: https://arxiv.org/pdf/1912.07768.pdf
Teaching with Commentaries: https://arxiv.org/pdf/2011.03037.pdf
Meta Pseudo Labels: https://arxiv.org/pdf/2003.10580.pdf
Reptile Chapters
0:00 Beginning
0:44 Reptile Motivation
5:05 Hyperparameters of Meta-Training
5:48 DataLoader
12:04 Visualizing Omniglot Samples
13:50 The Model
14:48 The Training Loop
19:52 Results
Watch video Few-Shot Learning with Reptile - Keras Code Examples online without registration, duration hours minute second in high quality. This video was added by user Connor Shorten 06 March 2021, don't forget to share it with your friends and acquaintances, it has been viewed on our site 10,676 once and liked it 265 people.