search

mattjj / pyhsmm

MIT License
550 stars 175 forks source link

Plot gaussian distributions for multidimensional data #66

Open dafnevk opened 8 years ago

dafnevk commented 8 years ago

The plotting functionality plots the first two dimensions of the observation data. For the gaussian observation distributions, it plots a projection on a random basis (in pybasicbayes ). Therefore, the ellipses for the distributions do not make sense with the plotted data points.

Would it be possible to also plot the ellipses on the first two dimensions? Even better would be if the plot function lets the user define which dimensions should be plotted.

danstowell commented 8 years ago

Hi - I added a simple pull request for this issue. Dipping my toe in the water of this package, feedback welcome.

mattjj commented 8 years ago

I forgot to come back to this issue; sorry, @dafnevk! Thanks for bumping it, @danstowell.

The plotting code is pretty rough around the edges. It sounds like I neglected to plot the ellipses in the right basis; thanks for catching it. The suggestion to have a user-specified plotting basis is also a good one. #68 is a decent initial fix, so I'll merge it now, but random projections might provide a more wholistic perspective on the data, and having user control is even better.

mattjj commented 8 years ago

I'm leaving this issue open so we can make better plotting code someday.