 |
 |
Hi > On 15 Feb 2020, at 03:06, Zhang Shaoshi <[log in to unmask]> wrote: > Hi FSL experts, > I have a question regarding how to perform tangent space parameterisation (using nets_tangentv.m) if the matrices are not necessarily positive definite (for example, partial correlation or Tikhonov regularised matrices). Those examples *should* always be SPD. But yes other cases might result in non-SPD. > Is it valid if I just set all the negative eigenvalues to some small positive values to force these matrices to be positive definite? In general, not necessarily the best thing to do, as that might change the matrix more than you wanted. However if the non-SPD-ness is very tiny (eg 1e-30 because of numerical precision) then yes that should be fine. But more generally you could use the following to get the nearest SPD matrix https://uk.mathworks.com/matlabcentral/fileexchange/42885-nearestspd Cheers. > > I am asking this question because I recently found this paper (Optimising network modelling methods for fMRI: https://www.biorxiv.org/content/10.1101/741595v3) and would like to try if tangent space parameterisation is helpful for our dataset as well. In the paper, it has been mentioned that 'Any eigenvalues close to zero were adjusted to a fixed, small non-zero value.' But how exactly is this done? Thank you. > > With the best regards, > Shaoshi > 15 Feb 2020 > > ######################################################################## > > To unsubscribe from the FSL list, click the following link: > https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1 --------------------------------------------------------------------------- Stephen M. Smith, Professor of Biomedical Engineering Head of Analysis, WIN (FMRIB) Oxford FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK +44 (0) 1865 610470 [log in to unmask] http://www.fmrib.ox.ac.uk/~steve <http://www.fmrib.ox.ac.uk/~steve> --------------------------------------------------------------------------- ######################################################################## To unsubscribe from the FSL list, click the following link: https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1