Direct Feedback Alignment

google-deepmind / dm-haiku

JAX-based neural network library

https://dm-haiku.readthedocs.io

Apache License 2.0

2.87k stars 233 forks source link

Direct Feedback Alignment #750

Open ferqui opened 11 months ago

ferqui commented 11 months ago

Hi, I'm training a recurrent network with haiku but I would like to apply Direct Feedback Alignment during training. The main issue I have encountered is how to send the output of the network into each layer gradient calculation. ¿Is there any option for that? Thank you