Martin R. Smith
Martin R. Smith
Current status: Not faster than phangorn.
in `app.R`, `thinnedTrees()` is naively defined as `as.integer(seq(keptRange()[1], keptRange()[2], by = 2 ^ input$thinTrees))` This assumes that trees have been loaded from a single file. Otherwise the numbers are garbage....
[Bogdanowicz and Krzysztof Giaro 2022](https://doi.org/10.1089/cmb.2022.0090) give a straightforward method; perhaps also consider comparing with their Jaccard-MC metrics.
https://journal.r-project.org/articles/RJ-2022-055/
- [x] Document how to colour trees by likelihood and contour the resultant space, per https://www.biorxiv.org/content/10.1101/2022.05.11.491507v1.full.pdf - [ ] Incorporate example from Wright & Lloyd 2020, to continue my polemic...
Could an implementation of [SIMD](https://www.intel.com/content/www/us/en/docs/intrinsics-guide/index.html) bring about c. 2x performance gains? - See also [1](https://stackoverflow.blog/2020/07/08/improving-performance-with-simd-intrinsics-in-three-use-cases/) - [2](http://www.cs.uu.nl/docs/vakken/magr/2017-2018/files/SIMD%20Tutorial.pdf)
Replicate in MapTrees() GUI by: - Selecting a file, and subsampling two batches of trees (I used best.tr from ms609/lobo) - Switching to display tab I see `ncol(distReference) == ncol(distLowDin)...
Currently calls phangorn's SPR.dist. We can improve performance and stability, and drop import of phangorn, by moving more of this from R to C.