scMerge now relies on BiocSingular for all SVD/PCA computations.
There is a reasonable integration with DelayedArray in addition to dgCMatrix.
Columns or rows with all zeroes are now automatically removed.
To be improved:
Due to various overheads, scMerge is still not as fast as it can be. Codes might need further optimisation in both memory usage and timing.
Updates:
scMerge
now relies onBiocSingular
for all SVD/PCA computations.DelayedArray
in addition todgCMatrix
.To be improved: Due to various overheads,
scMerge
is still not as fast as it can be. Codes might need further optimisation in both memory usage and timing.