Open asoplata opened 1 week ago
I think we have to accept that the GUI is meant for exploration and decide what features make most sense for teaching/exploratory purpose. Then export the data and continue analysis in HNN-core ... we'll soon have combinatorial problems to debug the more features we add to the GUI
That is a good point that we need to keep GUI additions to the minimum. We discussed this in the Dev meeting on 2024-11-15, and @ntolley and @dylansdaniels confirmed that the above is expected behavior.
In the short term, I will make a PR for this issue that adds a warning to the user. The warning will print to the main log in the event that the spectrogram is not generated due to the length of simulation being too low for the provided frequencies. This is why the spectrogram is not being shown in the default simulation of the GUI - they have clarified that the default GUI simulation is based off of the simulation used in the ERP tutorial, for which it is not useful to calculate a spectrogram.
In the meantime (for Milestone v0.4), @ntolley is going to work on changing plot_tfr_morlet
so that it automatically calculates its time x frequency analysis to be appropriate to the activity.
It seems that the hotfix I made to improve the spectrogram visuals for the Alpha/Beta tutorial here https://github.com/jonescompneurolab/hnn-core/pull/928 broke spectrogram display in the GUI when running the default simulation, at least for me. Before that commit it works, but after that commit, if I try to make a
Dipole-Spectrogram
figure after having run the default simulation that is loaded upon GUI start, I get this:This is almost certainly related to the length of time needed for analysis generation, which is why I needed to lengthen the time for the tests to pass here: https://github.com/jonescompneurolab/hnn-core/pull/928/commits/c0779ae733036308c541370d0932862bae48528f . How this apparently passed
pytest
runs, especiallytest_viz
andtest_gui
, I have no idea.The fast solution is simple: as we have done recently with the smoothing factor, we introduce a new hotfix to revert the Morlet cycle frequency divisor back to 8. Independent of the default, we should also add a widget so that users can customize the divisor itself.