I need to run a PyMC3 model in a loop to estimate/make predictions every month. How do you reset the Theano graph? I'm familiar with Tensorflow and I know this can be done, but googling doesn't seem to lead to any solutions. Alternatively, how are you meant to run a PyMC3 model in a loop?
Asked
Active
Viewed 141 times
0
-
Possibly related: [Incremental model update with PyMC3](https://stackoverflow.com/q/40870840/190597) – unutbu Oct 31 '18 at 18:27
-
Also: [How to store traces in a persistent format](https://docs.pymc.io/api/backends.html). – unutbu Oct 31 '18 at 18:29
-
Thanks for your reply. Looks like `theano.shared` is what I need, rather than resetting the graph every time. https://docs.pymc.io/notebooks/api_quickstart.html#4.-Posterior-Predictive-Sampling – swmfg Nov 01 '18 at 06:42