Consider replacing svd() with a faster function

[slowkow]

You might consider replacing stats::svd() with a faster function in order to support analysis of a greater number of cells.

Here's my benchmark:

[yulijia]

I tried a in-house dataset and replaced the stats::svd with irlba::irlba.

This method (phiclust) is not computationally efficient for large datasets.

out <- phiclust1(expr = mat, clusters = clusters)
Calculating values for cluster  1 
Dim:  24155 54994 
Calculating svd ... 
Error in La.svd(x, nu, nv) :
cannot allocate memory block of size 134217728 Tb