Hello! I’m writing to ask for advice/suggestions on creating a new operator class for operator VI in pymc3. Specifically, I’d like to use the squared L2 norm of the difference between the score function of the posterior and the variational approximation: \mathcal{L}(q,p)= \mathbb{E}_{\theta \sim …

Adding a new operator for an objective which involves gradients

ferrine April 1, 2019, 7:53pm 7

Do you have a demo notebook with your study? This would be a cool tutorial to extending OPVI in pymc3

Topic		Replies	Views
Variational API: meaning of parameters Questions	4	828	June 28, 2018
Regularization for AEVB Questions	2	619	June 26, 2018
ADVI giving NaN values to Elbo Questions	5	895	June 17, 2017
Using pm.DensityDist and customized likelihood with a black-box function Questions	10	6853	August 29, 2018
Perplexity of a model Questions from_github	21	2419	June 29, 2017