The VoxCeleb Speaker Recognition Challenge 2020 - Track 1 (Verification fully supervised, closed) Forum

Go back to competition Back to thread list Post in this thread

> Can we learn system fusion weights from the vox1cd dev set?

In the challenge FAQs, it is said:
"Q. For the CLOSED condition, can I use the validation set for training anything, eg. the PLDA parameters?
A. No, for the CLOSED condition you can use the validation set only to tune user-defined hyperparameters, eg. for example selecting which convolutional model to use."

But it is not clear whether we can learn system fusion weights by optimizing on the vox1cd set. Can the organizer clarify it?

Posted by: mandalorian @ Oct. 2, 2020, 4:51 p.m.
Post in this thread