ellipsoid_cut model shows poor correspondence

[akenmorris]

Here's what I see at the end of ellipsoid_cut. Notice there are swapped particles, etc.

Ellipsoid_Cut.mp4

[jadie1]

Will be fixed for release via issue #1133

[jadie1]

@HeavenlyBerserker After updating the ellipsoid data to only have one mode of variation I am still seeing bad correspondence. The shape model correctly predicts only one mode of variation but some of the particles are crossing over each other.
Can you look at the optimization parameters? The new dataset is called "ellipsoid_1mode_aligned". I have changed the ellipsoid_cut use case to use it on the "one_mode_ellipsoid" branch.

[HeavenlyBerserker]

@jadie1 I'll look at it on Thursday. Thanks for the note.

[sheryjoe]

@HeavenlyBerserker @jadie1 is this issue resolved?

[HeavenlyBerserker]

To the best of my knowledge, yes.

[jadie1]

The PR is using the old data still, it should be using ellipsoid_1mode_aligned. I am still seeing particles switch when the iterations are increased. I have tried a lot of different params and have had no luck so far...

[jadie1]

That was with 128 particles though. I will test with the new params on the one_mode_ellipsoid branch which uses the new data. If that looks good then we can close this

[jadie1]

Looks good, packing isn't super uniform but no particles are switching. Closing this now.