subsample: Feature-based subsampling of signals

In jiho/soundclass: Signal classifier for sounds

Description

Perform a partition of signals based on their characteristics and pick signals homogeneously across groups. This increases the probability of picking rare and different signals, which means that signals in the subsample are more representative of all signal types in the original data.

Usage

subsample(x, p, k = 10)

Arguments

x	data.frame/matrix of signals
p	proportion of the total number of signals to subsample, between 0 and 1
k	number of groups

Value

A data.frame containing picked and not-picked signals

See Also

plot.subsample, subsample_file

Examples

data(sirena)
head(sirena)
# The data contains:
# - signals as lines
# - signals features (V1->V8) and type as columns
# To take a feature-based subsample of signals, we need to remove
# the last column (i.e. the signal type)
sub <- subsample(sirena[,-ncol(sirena)], p=0.2)
head(sub)

jiho/soundclass documentation built on June 5, 2019, 10:10 p.m.