Joshua Taillon

@magnunor Just came across another need for this. Is there anything holding this back, currently? (given the 2.5 year gap...)

I see this is related to this deprecation: https://matplotlib.org/stable/api/prev_api_changes/api_changes_3.5.0.html#modification-of-axes-children-sublists

So the last example does not work on my system:   ``` $ pip list | grep -E "hyperspy|matplotlib" hyperspy 1.7.0 matplotlib 3.5.2 matplotlib-inline 0.1.3 ```

Ok, glad to know I'm not crazy! I'm curious why one type of marker works and the other doesn't. Maybe it's an underlying bug in matplotlib? I'd have to look...

Just commenting an opinion that (as a user), if I specify interpreter versions that the project might use via `pyenv local X.Y.Z x.y.z ...`, I would naively expect that `pyenv...

Do you see any negative consequences of adding an infinitesimally-small value to channels that are uniformly-zero before proceeding with the MCR algorithm? I suppose it could preferentially bias the solution...

@CCampJr I mean portions of the spectra. i.e. if you have an *x* *y* spectrum image map of spectra, if one of the channels is zero-valued at every *x* and...

*I just saw you responded, but I'll send this comment through anyway...* Yes, zero across the whole image. Here's an example of what I mean (plotted within HyperSpy). This is...

I don't believe HyperSpy is doing anything unusual, but I will try to reproduce using only `pyMCR` and not HyperSpy to make sure I see the same behavior... Give me...

Unfortunately, no they're floats. It's required by the SVD that we run prior to feeding those in as initial conditions to pyMCR