William Hicks
William Hicks
This is a known issue, with a fix available on an internal branch up for public release.
Unfortunately, it's now failing earlier (at the compute_many_descriptors stage), so I can't tell if this issue has been resolved. I'll look into this new error and open a new issue...
I'd be happy to have a crack at this if it's still of interest. I didn't see much on the [dev/vptree-nn-index](https://github.com/Kitware/SMQTK/tree/dev/vptree-nn-index) branch, so a few questions: 1. Was there an...
Okay, great! Responding point-by-point: 1. I'll go ahead and start doing my own implementation until/unless you can put your local implementations out in some form. If you do put it...
Makes sense, thanks! I'll get to work.
Great, thanks! I'll take a look at it over the weekend and compare against what I've got so far.
Sorry, got my own wall of text for you this time! After going through the Yianilos paper again, I think the poor performance with the Hamming distance metric is inherent...
Okay, thanks! In that case, it may still be worth it. I'll run some tests at around those sizes and let you know what I find out.
I ran a bunch of tests over the weekend with your VP tree implementation (integrated into [this fork](https://github.com/wphicks/SMQTK/commits/vp-tree-integration)). A few results: 1. All variants of VP trees start to outperform...
Yes, I have a couple scripts that I used to generate a whole bunch of plots for different dimensions and numbers of indexed vectors. I was planning on putting the...