How to implement the negative binomial likely hood function in tensorflow and use it as the loss function to train an RNN?

Asked Jan 13 '19 at 05:23

Active Feb 19 '20 at 11:29

Viewed 467 times

I am trying to predict the mean and variance of a random variable in each step of a recurrent neural network. I need to train this model by using negative binomial log-likelihood as the loss function where the model learns to output mean and variance and each time step which defines the distribution. Can someone please help me with implementing this loss function in tensorflow.

asked Jan 13 '19 at 05:23

Shamane Siriwardhana

3,951
6
33
73

Can you elaborate it please? and also implementing custom loss function is pretty straightforward https://stackoverflow.com/questions/34875944/how-to-write-a-custom-loss-function-in-tensorflow – ARAT Jan 14 '19 at 05:25
Can you please refer to the loss function mention in - https://arxiv.org/pdf/1704.04110.pdf . Page 4 equation 2 – Shamane Siriwardhana Jan 14 '19 at 05:37

How to implement the negative binomial likely hood function in tensorflow and use it as the loss function to train an RNN?

0 Answers0