H5matref: obtain an HDF5 dataset reference suitable for handling as...

In BiocSklearn: interface to python sklearn via Rstudio reticulate

Description

obtain an HDF5 dataset reference suitable for handling as numpy matrix

Usage

H5matref(filename, dsname = "assay001")

Arguments

filename	a pathname to an HDF5 file
dsname	internal name of HDF5 matrix to use, defaults to 'assay001'

Value

instance of (S3) "h5py._hl.dataset.Dataset"

Note

This should only be used with persistent environment discipline of basilisk. Additional support is planned in Bioc 3.12.

Examples

## Not run: 
fn = system.file("ban_6_17/assays.h5", package="BiocSklearn")
ban = H5matref(fn)
ban
proc = basilisk::basiliskStart(bsklenv)
basilisk::basiliskRun(proc, function() {
 np = import("numpy", convert=FALSE) # ensure
 print(ban$shape)
 print(np$take(ban, 0:3, 0L))
 fullpca = skPCA(ban)
 dim(getTransformed(fullpca))
 ta = np$take
 })
basilisk::basiliskStop(proc)

## End(Not run)
# project samples
## Not run:   # on celaya2 this code throws errors, and
#  I have seen
# .../lib/python2.7/site-packages/sklearn/decomposition/incremental_pca.py:271: RuntimeWarning: Mean of empty slice.
#   explained_variance[self.n_components_:].mean()
# .../lib/python2.7/site-packages/numpy/core/_methods.py:85: RuntimeWarning: invalid value encountered in double_scalars
#   ret = ret.dtype.type(ret / rcount)
ta(ban, 0:20, 0L)$shape
st = skPartialPCA_step(ta(ban, 0:20, 0L))
st = skPartialPCA_step(ta(ban, 21:40, 0L), obj=st)
st = skPartialPCA_step(ta(ban, 41:63, 0L), obj=st)
oo = st$transform(ban)
dim(oo)
cor(oo[,1:4], getTransformed(fullpca)[,1:4])

## End(Not run) # so blocking this part of example for now

BiocSklearn documentation built on Nov. 8, 2020, 7:52 p.m.