Faster Nested Monte Carlo Estimator for EIG

ricardoV94 · July 23, 2024, 9:07pm

You could try to use vectorize_graph to create a batched version of lp_cond that handles all the values of observed_y at once.

github.com

pymc-devs/pytensor/blob/d45546084c1041af63cfcca43ab8d32ff04a2472/pytensor/graph/replace.py#L238


      
          ) -> Variable: ...
          
          
          @overload
          def vectorize_graph(
              outputs: Sequence[Variable],
              replace: Mapping[Variable, Variable],
          ) -> Sequence[Variable]: ...
          
          
          def vectorize_graph(
              outputs: Variable | Sequence[Variable],
              replace: Mapping[Variable, Variable],
          ) -> Variable | Sequence[Variable]:
              """Vectorize outputs graph given mapping from old variables to expanded counterparts version.
          
              Expanded dimensions must be on the left. Behavior is similar to the functional `numpy.vectorize`.
          
              Examples
              --------
              .. code-block:: python

Recommend trying with the latest version of PyMC

Topic		Replies	Views
Draw_values() speed/scaling with transformed variables Questions	9	2006	November 7, 2019
How to calculate log posterior of a GP over a trace in pymc3 Questions	14	2286	February 2, 2018
Help with Multivariate Normal with nested prior information Questions	4	1618	January 7, 2018
Estimating probability of data point using inferred posterior Questions	7	775	April 30, 2018
Very parallel MCMC sampling Development	7	4126	August 22, 2019

Faster Nested Monte Carlo Estimator for EIG

Related topics