Uncoupling node dynamics

[NeuroLife]

Hi,

I was reading about different ways to reduce the dimensionality of the parameter space to estimate when model fitting and I came across a paper (1) that describes a method to uncouple the dynamics of the different nodes within a simulation with TVB by subtracting the excitatory mean-field potentials from coupled nodes for each nodes. I was wondering if this was implemented or if it has to be done manually in some way. 

I would like to investigate the potential of this approach for my project but I have relatively limited knowledge of what happens "under the hood" of TVB, so if it has to be done manually I would appreciate a few pointers in terms of where to look within the TVB code to be able to implement this.

(1) Ritter, P., Schirner, M., McIntosh, A. R., & Jirsa, V. K. (2013). The virtual brain integrates computational modeling and multimodal neuroimaging. 

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3696923/ 

Thank you for your time,

Have a nice day.

Dominic Boutet

[WOODMAN Michael]

hi

 

The idea of that technique is that the EEG source estimate represents the local activity driven by inputs from other nodes.  If there are no delays, then you know the inputs are just the activity at other nodes at the previous time point, weighted by the connectome, so you can remove this contribution to the difference from the previous to current time points, to obtain the ‘uncoupled’ local dynamics.  The process would then be iterated over all time steps.  This then provides an estimate of the uncoupled local dynamics per node, which is easier to fit parameters to.  If you have delays, it just complicates the algorithm a bit, but the concept is the same.

 

The TVB software doesn’t implement this algorithm, though it would be useful to generate simulated data and test the algorithm to check it works for your purposes:  it would only work for coupling in a linear regime, for instance, which isn’t true of all models.  Another caveat is that, from a statistical estimator perspective, it hinges on low error as it iterates through time, as do the subsequent parameter estimates.  More recent techniques which use Bayesian autoregressive approaches do not have the same issue, and especially those using gradient-based algorithms handle large parameter spaces quite well, given appropriate regularizations.

 

cheers,

Marmaduke

--
You received this message because you are subscribed to the Google Groups "TVB Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to tvb-users+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/tvb-users/f4ee9d8a-2fdc-4d5a-817d-74fb018f54c2n%40googlegroups.com.

[NeuroLife]

Ok, I will look into that.

Thank you!

Best.

Dominic Boutet