pymc3: finding lowest misfit using MCMC

by Alexander Van Ballaer   Last Updated August 13, 2019 20:19 PM - source

I have made a function that calculates the misfit(squared difference) between an observed signal at a target station and reconstructed signal from 10 arbitrary stations chosen from a dataset. There are a total of 105000 stations in the dataset. See the code below for clarity.

ds1 = ASDFDataSet("output1_custom/receivers.h5") #dataset from earthquake simulation

Target = 500 #enter target station
Total_trace = [Target,np.arrange(10)] #combination of target and reconstruction stations (needed as input for loaddata function)
misfit=loaddata(Target,Total_trace,ds1) #loaddata function calculates the misfit

I was wondering if it is possible to find the combination of 10 stations (out of the 105000) that minimizes this misfit, with MCMC. I am brand new to the pymc3 package, and none of the tutorials appear to be relevant to what I am attempting.

Any comments/criticism is welcome!



Related Questions


Why doesn't k-means give the global minimum?

Updated March 01, 2019 13:19 PM

Show that $nX_{(1)}$ is not consistent

Updated June 18, 2019 14:19 PM

Poisson as hidden variable fails in PyMC3

Updated March 02, 2017 05:19 AM