search

google / uis-rnn

This is the library for the Unbounded Interleaved-State Recurrent Neural Network (UIS-RNN) algorithm, corresponding to the paper Fully Supervised Speaker Diarization.
https://arxiv.org/abs/1810.04719
Apache License 2.0
1.55k stars 320 forks source link

[Question] how to relate sigma2_prior_loss() with the paper #52

Closed jprobichaud closed 5 years ago

jprobichaud commented 5 years ago

First, thanks for sharing this implementation.

We're trying to understand the core of the training methods with the paper' equations. We're able to make sense of most of it, but the sigma2_prior_loss() in loss_func.py escape our understanding.

Could someone provide some explanation for it? How to derive that loss from the paper?

Thanks in advance!

wq2012 commented 5 years ago

Please see issue #10 for the answers.

Regularization loss terms were not discussed in paper since it had a 4-page limit.