pgen.1010657.s004.pdf (127.96 kB)

Loss function value trajectories calculated on training and validation data for versions of IntroUNET with increasing sample sizes (32, 64, and 128 individuals per subpopulation), training set sizes (1000, 10000, and 100000 alignments), and window sizes (64, 128, and 256 polymorphisms).

journal contribution

posted on 2024-02-20, 18:26 authored by Dylan D. Ray, Lex Flagel, Daniel R. Schrider

All tests were calculated on simulated examples of the simple bidirectional scenario described in the Methods.

(PDF)

History

Usage metrics

Categories

Keywords

world scenarios underscores trained neural network much higher frequencies insufficient — ideally div >< p deep neural network deep learning approaches deep learning algorithm significant fitness advantage closely related species recovering introgressed haplotypes image classification problem harbor introgressed loci region previously shown use simulated data population genetic alignment identifying introgressed alleles use ="population alignment one species fitness effects correctly identifying introgressed material introgressed alleles image representation image belongs widespread phenomenon typically neutral typically confined sometimes confer semantic segmentation real data readily extended purifying selection potential relevance performing comparably numerous methods identify regions identify alleles highly effective highly accurate growing body gene flow full extent evidence suggests drosophila close relative challenging real analysis reveals >, showing

Licence

Exports

RefWorks

BibTeX

Ref. manager

Endnote

DataCite

NLM

DC