search

katerakelly / pytorch-maml

PyTorch implementation of MAML: https://arxiv.org/abs/1703.03400
MIT License
553 stars 129 forks source link

Question about meta-update in Algorithm 2 mentioned in the paper. #2

Closed dingli-dean closed 6 years ago

dingli-dean commented 6 years ago

In the algorithm 2 mentioned in this paper,we have computed adapted parameters θ' in step 7，and I think we have updated the model for task T_i. What does the meta-update mean?(in step 8)

katerakelly commented 6 years ago

In the meta-update step, the gradient is taken of the loss computed with the adapted parameters, with respect to the original parameters. So the meta-update step adjusts the parameters such that when they are adapted to a task, the loss is minimized.

Please reserve Github issues for bugs, feature requests, etc. You can ask questions about machine learning on sites such as Cross Validated, or Reddit. Questions about PyTorch should be asked on the PyTorch Forum.

dingli-dean commented 6 years ago

Thank you very much.