search

BRML / climin

Optimizers for machine learning
Other
180 stars 66 forks source link

About updates in adadelta #38

Open MilesQLi opened 8 years ago

MilesQLi commented 8 years ago

I think in the original paper of adadelta, there is no step_rate.

f0k commented 8 years ago

That's right. Feel free to add a note to the documentation. In Lasagne, we have (a bit sloppy): "In the paper, no learning rate is considered (so learning_rate=1.0). Probably best to keep it at this value."