dokato
dokato
One idea is to replace the weighted scores between the closest dotprops, similarly to alpha. The weights can be a distance from the soma with a pairwise score: ``` 1...
Another idea is using topological sort of each neuron and then scoring neurons in that order (hence we don’t use k-NN). This may trigger problems as shown in the figure...
@schlegelp yeap this has been done already ;) still testing what metric works best for weighting though...
Not sure, I tested that myself a few days ago: https://github.com/dokato/nat.nblast/tree/topo
Nope, so far in all cases, I'm using default `smat.fcwb`. Is it easy to generate a new scoring matrix? If so, I can do that quickly with new metrics.
Thanks for reporting this. Indeed qmd files were not thoroughly tested with todor yet, but I'll have a closer look.
I can help with review and also added a few changes, but here are some general comments. - Some sentences read difficult (I recomend using [Hemingway app](https://hemingwayapp.com/) to check for...
Okay, I gave it another pass, you may have a look. I decided to split Open source efforts and Quality Assurance. Still tempted by moving "Related Work" before "Methods" but...
Hey @KennethEnevoldsen I'd like to merge also this dataset in #773. 3 reasons: a) we don't seem to have brazilian dialect represeted, b) multilabel task doesn't have large language coverage...
Yes, totally, I'll help!