search

tobiscode / disstans

Package repository for the Decomposition and Inference of Sources through Spatiotemporal Analysis of Network Signals (DISSTANS) toolbox.
https://tobiscode.github.io/disstans/
GNU General Public License v3.0
22 stars 7 forks source link

Is it possible to update the code to treat the tau parameter as one that can be solved for within the model? #10

Closed GEOSCIENCELXG closed 2 hours ago

GEOSCIENCELXG commented 2 hours ago

In the current version, we need to provide a tau value for the logarithmic model; however, tau is a crucial parameter that we would like to estimate instead.

Is it possible to update the code to treat the tau parameter as one that can be solved for within the model?

Cheers

tobiscode commented 2 hours ago

Hi, estimating tau would make the problem heavily non-linear, and is therefore definitely out of the scope and technical possibilities of DISSTANS. An alternative that may or may not work for you in your circumstance (it does for mine), is to use the option to have multiple taus, and estimate coefficients for each of these. The trick here is that a given tau can be well approximated by a weighted sum of multiple other taus. (Also: rarely does nature settle on a single tau anyway.) Check out the documentation for both Logarithmic and Exponential, they both have this option. Note: you will probably have to regularize. Cheers