Parameter estimation

57 views
Skip to first unread message

NeuroLife

unread,
Jun 17, 2021, 4:52:02 PM6/17/21
to TVB Users
Hi,

I was wondering if anyone had any insight on how to do parameter estimation in TVB from real life data (ideally PSDs). So far, I have tried a simple python minimization approach, but it proved to be computationally expensive and not really reasonable within my current timeframe. I have also looked into deep learning based regression, but it is quite difficult to use (especially since I am very new to deep learning) and I don't know if it could provide accurate estimations.

So, if anyone could provide insight or send papers that may contain insight on this, I would appreciate greatly.

Thank you for your time,
Have a nice day.
Dominic Boutet

WOODMAN Michael

unread,
Jun 18, 2021, 2:45:50 AM6/18/21
to TVB Users

Hi


TVB doesn't provide estimation for the moment.  You can use TVB with gradient free and likelihood free techniques like SBI (https://www.mackelab.org/sbi/) or write the model equations as components in a framework like Pyro.  You can find examples in the BVEP repo,


https://github.com/ins-amu/BVEP/tree/master/SBI

https://github.com/ins-amu/BVEP/tree/pyro-vi/PyroVI/vep


cheers,

Marmaduke


From: tvb-...@googlegroups.com <tvb-...@googlegroups.com> on behalf of NeuroLife <domin...@hotmail.com>
Sent: Thursday, June 17, 2021 10:52:02 PM
To: TVB Users
Subject: [TVB] Parameter estimation
 
--
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/6329c578-6924-4583-a1cb-eb75de1ab02bn%40googlegroups.com.

NeuroLife

unread,
Jun 18, 2021, 2:02:13 PM6/18/21
to TVB Users
Hi,

Thank you this is very helpful!

Have a nice day.
Dominic Boutet

NeuroLife

unread,
Jun 23, 2021, 5:56:52 PM6/23/21
to TVB Users

Hi,

I am using the SBI tool from mackelab for parameter estimation and I was wondering if using a specific feature of the data as input would have better or worse chances at providing efficient fitting (e.g. full timeseries vs PSD). 
I would ideally want the simulated data resulting from the estimated parameters to fit real life data across multiple features, not just the one used as input for the estimation. Although, despite my limited of knowledge of the subject I have a feeling that this may be a bit too much to ask from current estimation approaches considering computational limitations. 

If anyone has insight or recommended readings on this subject it would be greatly appreciated.

Thank you for your time.
Dominic Boutet
Le vendredi 18 juin 2021 à 02 h 45 min 50 s UTC-4, marmaduke.woodman a écrit :

WOODMAN Michael

unread,
Jun 24, 2021, 1:05:15 AM6/24/21
to TVB Users

Hi


I'd suggest searching for "feature engineering" for the general principles.  After that you'd want to choose features relevant to this domain such as FC and PSD, etc.  Multiple can be used with SBI, but I don't expect it to generalize  (reproduce features not provided as input).


cheers,

Marmaduke

Sent: Wednesday, June 23, 2021 11:56:52 PM
To: TVB Users
Subject: Re: [TVB] Parameter estimation
 
Reply all
Reply to author
Forward
0 new messages