question about dataset - Githubissues

openai / supervised-reptile

Code for the paper "On First-Order Meta-Learning Algorithms"

https://arxiv.org/abs/1803.02999

MIT License

996 stars 214 forks source link

question about dataset #29

Open YuBeomGon opened 4 years ago

YuBeomGon commented 4 years ago

In the paper, I found this sentence, Unlike FOMAML, Reptile doesn’t need a training-test split for each task, which may make it a more natural choice in certain settings (https://arxiv.org/pdf/1803.02999.pdf)

it means I dont need to split data?(support set and query set in each episode in train data set) Am I right?

unixpickle commented 4 years ago

I think you are interpreting the sentence correctly. Basically, Reptile doesn't require you to split up each mini-dataset during meta-learning. It seems that FOMAML works mostly because there's a separate mini-test-set that has a useful gradient after the model has fully overfit to the mini-training-set.