skIncrPCA_h5: demo of HDF5 processing with incremental...

In vjcitn/BiocSklearn: interface to python sklearn via Rstudio reticulate

demo of HDF5 processing with incremental PCA/batch_size/fit_transform

Description

demo of HDF5 processing with incremental PCA/batch_size/fit_transform

Usage

skIncrPCA_h5(fn, dsname = "assay001", n_components, chunk.size = 10L)

Arguments

fn	character(1) path to HDF5 file
dsname	character(1) name of dataset within HDF5 file, assumed to be 2-dimensional array
n_components	numeric(1) passed to IncrementalPCA
chunk.size	numeric(1) passed to IncrementalPCA as batch_size

Note

Here we use IncrementalPCA$fit_transform and let python take care of chunk retrieval. skIncrPartialPCA acquires chunks from R matrix and uses IncrementalPCA$partial_fit.

Examples

if (interactive()) {
 fn = system.file("hdf5/irmatt.h5", package="BiocSklearn") # 'transposed' relative to R iris
 dem = skIncrPCA_h5(fn, n_components=3L, dsname="tquants")
 dem
 head(getTransformed(dem))
}

vjcitn/BiocSklearn documentation built on Feb. 4, 2024, 5:12 a.m.