Hello, I am carrying out Bayesian regression and want to predict on the out-of-sample data. For this, I’m using pm.Data() command to set the input training data. I notice that I encounter dimension mismatch in the likelihood function if I use theano shared variables. However, this is not the case …

The problem is that broadcastable. You have to tell Theano that the first dimension of data is broadcastable (size 1). I don’t know if in your version of PyMC you can already pass broadcastable=(True, False) when creating x1 and x2. You can in the new beta version of PyMC for sure, and if not in you…

Dimension mismatch in the normal likelihood with theano variables

Questions version agnostic

cluhmann March 9, 2022, 4:40pm 8

I might suggest opening a new topic (with this topic marked as “solved”, people may be less likely to read/respond).

1 Like

Topic		Replies	Views
Shape mismatch using theano shared variable Questions	20	1676	February 1, 2019
Shapes of shared variables Questions theano	3	987	June 29, 2018
Test_value shape errors with theano.shared Questions	13	2586	March 27, 2018
Theano error with element wise multiplication Questions theano	5	2525	July 12, 2017
Predicting on hold-out data，Error：Input dimension mis-match. (input[1].shape[0] = 126, input[2].shape[0] = 3) Questions	8	3070	December 8, 2017

Dimension mismatch in the normal likelihood with theano variables

Related topics