I have a model with a mix of continuous and discrete variables. From the documentation, it looks like pymc3 can handle that by using NUTS for the continuous variables and metropolis-hastings for the discrete ones. def Matern52(X1, X2, alpha, rho): L2 = np.sum(X1 ** 2, axis=1, keepdims=True) + …

There shape error is a bit odd, but reparameterize it into something like below should work: Xshare = theano.shared(X) # cast into theano so you can index it using pymc3/theano tensor with pm.Model() as model: ... # use categorical because Metropolis gives invalid proposal value for Discret…

TypeError in NUTS

Questions

yangkky May 9, 2018, 7:02am 4

Using your method, it runs, probs is always zero in the trace – it never explores.

Topic		Replies	Views
Dimension mismatch in `NUTS` and `metropolis` Questions	3	927	August 21, 2017
PyMC3 on AWS EC2 Questions	15	1809	December 12, 2017
Issue with DiscreteUniform distribution Questions	3	961	January 30, 2018
Indexing using a free random variable Questions	26	6452	October 16, 2019
When ever I tried running NUTS sampling my notebook gets disconnected from kernel Questions	4	1872	July 7, 2018

TypeError in NUTS

Related topics