A question about the inner gradient calculation

cbfinn / maml

Code for "Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks"

MIT License

2.55k stars 604 forks source link

A question about the inner gradient calculation #39

Closed Codelegant92 closed 6 years ago

Codelegant92 commented 6 years ago

Hi Chelsea! In you paper, during meta training process, the inner gradients(theta') are calculated independently for different tasks within one batch. Then the updated parameters are applied to the new samples to calculate the outer gradients(theta). However, in your code maml.py - task_metalearn(), I didn't see independent calculation of the inner gradients for different tasks within a batch. Can I ask why?

cbfinn commented 6 years ago

The map_fn parallelizes across tasks in the batch. task_metalearn corresponds computation for a single task. Hence, the gradient computation is performed separately for each task.