deepmind
What regularizers do you use when training?
I noticed that in Conv3D or BatchNorm modules, the default regularizers are None. Do you use regularizers in Conv3d or BatchNorm when training? If so, do you use L1 or L2 regularizers?
Hi Junhui,
we used l2 regularizers but typically quite small (10^-7 or lower) on Kinetics as the model does not overfit too much there.
Joao
On Sat, Jul 21, 2018 at 8:10 AM, Junhui Wu [email protected] wrote:
I noticed that in Conv3D or BatchNorm modules, the default regularizers are None. Do you use regularizers in Conv3d or BatchNorm when training? If so, do you use L1 or L2 regularizers?
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/deepmind/kinetics-i3d/issues/27, or mute the thread https://github.com/notifications/unsubscribe-auth/AO6qaqr7CUi4sUpoElvYUakaole4PieZks5uItP5gaJpZM4VZhpw .
