Hi, I want to implement a new sampler and need the banana-shaped distribution as a test case. Here I found an implementation for pymc, but I can’t transfer it to pymc3. This is how I can transform samples from a multivariate normal into the banana-shape: def banana(X, b=0.03): """Twist the s…

[image] michaelosthege: I want to implement a new sampler and need the banana-shaped distribution as a test case. You can define a banana-shaped function and wrap it using pm.Potential. For example, defining a Rosenbrock function : import numpy as np import pymc3 as pm import theano.tensor as…

Sampling the Banana-shaped distribution

Questions

junpenglao July 23, 2018, 7:35pm 12

Right, lambda function sometimes is non pickle-able, in these cases, sample with cores=1 should resolved the issue.

1 Like

Topic		Replies	Views
When ever I tried running NUTS sampling my notebook gets disconnected from kernel Questions	4	1872	July 7, 2018
Sample method of pymc3 Questions	4	685	July 15, 2018
NUT sampler stuck under windows with njobs>1 Questions	13	4376	April 27, 2020
NUTS sampling doesn't work Questions	9	3091	December 21, 2020
Switch point Metropolis tuning Questions	25	5000	May 19, 2018

Sampling the Banana-shaped distribution

Related topics