2

I am trying to combine cvxopt (an optimization solver) and PyMC3 (a sampler) to solve convex stochastic optimization problems.

I have tried to use the following code (PyMC, version 2) as a baseline. But I couldn't make it work.

Stochastic Optimization in Python

First, when I tried to just replacing pymc with pymc3, I got the following error:

TypeError: No model on context stack, which is needed to instantiate distributions. Add variable inside a 'with model:' block, or use the '.dist' syntax for a standalone distribution.

Then I add my code inside a "with" statement as follows, and got an "AttributeError".

AttributeError: module 'pymc3' has no attribute 'deterministic'

import numpy as np, pymc3, cvxopt as co

# suppress cvxopt solver output, since it will be inside MCMC loop
co.solvers.options['show_progress'] = False


with pymc3.Model() as model:
    c1 = pymc3.Normal('c1', mu=-4, tau=0.5**-2)

    @pm.deterministic
    def x(c1=c1):
        c = co.matrix([float(c1), -5.])
        G = co.matrix([[2., 1., -1., 0.], [1., 2., 0., -1.]])
        h = co.matrix([3., 3., 0., 0.])
        sol = co.solvers.lp(c, G, h)
        solution = np.array(sol['x'],dtype=float).flatten()
        return solution

    m = pymc3.MCMC(dict(c1=c1, x=x))
    m.sample(20000, 10000, 10)

I decided to comment "#@pymc3.deterministic", and then I got the following error:

AttributeError: module 'pymc3' has no attribute 'MCMC'

So, I am stuck at this point. I don't know what would be the equivalent of calling MCMC using pymc3. Everything I tried so far is giving me the following message:

c = co.matrix([float(c1), -5.]): TypeError: float() argument must be a string or a number, not 'FreeRV'

Let me know if you need more details.

Nelson M. Fernandes
  • 21
  • 2
  • What do you mean by *couldn't make it work*? Please include your code as a an [MCVE](https://stackoverflow.com/help/mcve), the desired behavior, and the error you're facing. – Nino Filiu Jan 31 '19 at 14:55
  • @NinoFiliu Thanks for the reply. I tried to add more details in the post. Let me know if it is clearer now. – Nelson M. Fernandes Feb 01 '19 at 15:21

0 Answers0