SOM.cluster: Function to apply a pipeline of deconstruction, learning,...

In DimitriF/DLC: Deep Layer Chromatography

Description

Function to apply a pipeline of deconstruction, learning, reconstruction on a picture

Usage

SOM.cluster(data, margin, transform = c(F), grid.x = c(3),
  grid.y = c(3), topo = c("hexagonal"), toroidal = c(F),
  rlen = c(100), alpha.1 = c(0.05), alpha.2 = c(0.01),
  radius = c("Default"), action = c("reconstruct"),
  count.successif.1 = rep(0, length(margin)),
  count.successif.2 = rep(100, length(margin)), use = length(margin))

Arguments

data	a 3D array, create by the function f.read.image
margin	the margin to apply the deconstruction
transform	should the array be transformed after deconstruction (interesting value are: margin=3-transform=F and margin=2-transform=T)
grid.x	x width of the kohonen map
grid.y	y width of the kohonen map
topo	topo of the kohonen map
toroidal	should the map be toroidal
rlen	number of iteration
alpha.1	learning rate begining
alpha.2	learning rate end
radius	radius of the neighbourhood
action	what the process should return for the next step, choices are: reconstruct, evolve, clusterize, original, original_noirci.
count.successif.1	for original_noirci what is the minimum of successive sample to be consider as sample and note noise or interband
count.successif.2	for original_noirci what is the maximum of successive sample to be consider as sample and note noise or interband
use	use to skip a process

Author(s)

Dimitri Fichou

Examples

data <- f.read.image('www/rTLC_demopicture.JPG',format='jpeg',native=F) %>% redim.array(256)
model <- SOM.cluster(data,margin=c(3),transform=F,action='evolve')
str(model)
model$data.recon[[1]][,,1] %>% raster

DimitriF/DLC documentation built on Oct. 14, 2020, 4:33 p.m.