colLogSumExps: Accurately calculates the logarithm of the sum of...

colLogSumExps,DelayedMatrix-methodR Documentation

Accurately calculates the logarithm of the sum of exponentials for each row (column) of a matrix-like object

Description

Accurately calculates the logarithm of the sum of exponentials for each row (column) of a matrix-like object.

Usage

## S4 method for signature 'DelayedMatrix'
colLogSumExps(
  lx,
  rows = NULL,
  cols = NULL,
  na.rm = FALSE,
  force_block_processing = FALSE,
  ...,
  useNames = TRUE
)

## S4 method for signature 'DelayedMatrix'
rowLogSumExps(
  lx,
  rows = NULL,
  cols = NULL,
  na.rm = FALSE,
  force_block_processing = FALSE,
  ...,
  useNames = TRUE
)

Arguments

lx

A NxK DelayedMatrix. Typically, lx are log(x) values.

rows, cols

A vector indicating the subset (and/or columns) to operate over. If NULL, no subsetting is done.

na.rm

If TRUE, missing values (NA or NaN) are omitted from the calculations.

force_block_processing

FALSE (the default) means that a seed-aware, optimised method is used (if available). This can be overridden to use the general block-processing strategy by setting this to TRUE (typically not advised). The block-processing strategy loads one or more (depending on ⁠\link[DelayedArray]{getAutoBlockSize}()⁠) columns (colFoo()) or rows (rowFoo()) into memory as an ordinary base::array.

...

Additional arguments passed to specific methods.

useNames

If TRUE (default), names attributes of result are set. Else if FALSE, no naming support is done.

Details

The S4 methods for x of type matrix, array, table, or numeric call matrixStats::rowLogSumExps / matrixStats::colLogSumExps.

Value

Returns a numeric vector of length N (K).

Author(s)

Peter Hickey

See Also

  • matrixStats::rowLogSumExps() and matrixStats::colLogSumExps() which are used when the input is a matrix or numeric vector.

  • rowSums2()

Examples

x <- DelayedArray(matrix(runif(10), ncol = 2))
colLogSumExps(log(x))
rowLogSumExps(log(x))

PeteHaitch/DelayedMatrixStats documentation built on Nov. 9, 2024, 10:53 a.m.