Refuses rather than improvises when the feature sets differ: a model trained on 11 markers cannot place cells described by 9, and quietly filling the gap with zeros would produce coordinates that look plausible and mean nothing.
Arguments
- saved
A model loaded by
load_umap_model().- mat
Numeric matrix, cells x features.
- n_threads
Number of threads. Default
1L.- transform
this run's transform; where it differs from the one the model stored, each feature is carried onto the stored one first.
