Open valentinwust opened 1 month ago
Hi @valentinwust
Thanks for reaching out and apologies for the delayed response, I was out last week. I would be interested in this, though Torch is a pretty heavy dependency. How about we add this to the flowkit-extras package, which I by the issue you submitted there that you also found.
FlowKit-extras started as somewhat of a catch-all package for experimental ideas or things that don't fit nicely in the main packages (FlowIO, FlowUtils, and FlowKit). It's a playground where I don't worry about new dependencies so much. It would be nice to utilize the FlowKit Sample class for inputting the data to such a function. How does that sound?
Thanks again, Scott
Hello,
panelbuilder has some nice unmixing options that go beyond OLS. I haven't tried too much of it, but especially removing specific spectra from the data (e.g. an autofluorescence from single stain cells) seems interesting. I'll probably reproduce a lot of what they do in Python anyway, would you be interested in me adding that functionality to the package? Their special unmixing would add a torch dependency, but that could be treated as an optional dependency. (I could technically also implement it in pure numpy, but GPU support is pretty useful if the data get large.)
Greetings Valentin