Closed djanloo closed 1 month ago
Numerical noise is highly reduced if Jansen-Rit population potential is plotted instead of rate. This probably means that the error is introduced in the conversion.
A huge part of the numerical error was introduced while streaming to the output file. Setting the precision to the max number of significant digits solved the issue.
Rate (timestep = 2ms):
This was an output bug. Setting the parameters of the oscillators to double
does not substantially affect the precision. Parameters are kept to be float
.
The noise in interpolation between timesteps seems to build up over time. This should not happen since the polynomials are fixed.![bug_NCE8](https://github.com/djanloo/quilt/assets/89815653/3a76468e-76c9-4692-ba76-80c9d1eee5a0)