In caffe cifar10 example, there're no weight decay in some convolutional layers.

29 views

Skip to first unread message

unread,

Aug 8, 2015, 11:08:18 PM8/8/15

to Caffe Users

And even no learning rate in "conv3", how is this ? why is the absence of decay_mult and lr_mult in weights layer ?

Reply all

Reply to author

Forward

0 new messages