mxNoisifyPackage

The task

The problem to solve is to add some random noise to a matrix, to get K slightly different matrices from the original one.

Inputs:

• inputPath: a path to a .csv file with a matrix: a header row and M x N numbers
• K: how many different random matrices we need
• p: the amount of noise to be added (uniform random variable should be generated for each position in the matrix independently between -p and +p)
• outPath: a directory to write the new matrices into in csv format, and a new header that should contain the p value and the random seed that was used.

Substeps (suggested functions):

• read in matrix data from a csv file
• add random uniform noise to a matrix with any dimensions (random seed should be a parameter)
• a loop to repeat the random noise generation K times
• write a noisified matrix to a path, with an extra row as a first row starting with '#', then the original header row, then the noisified data

How to run:

./generateRandMxes.sh

Hints

To read in a matrix:

inDataTable <- read.table(inputMxPath, skip=1)   # skipping header row
inputMx <- as.matrix(inDataTable)

To write a matrix into a file:

myWriter <- file(outFilename, "w")
write("# First row of file", myWriter)
write("# Second row of file", myWriter, append = TRUE)
write.table(myNoisyMatrix, file=outFilename, quote=FALSE, append = TRUE, col.names = FALSE)
close(myWriter)

To set the random seed

set.seed(23)

To generate a vector of 15 random numbers between -2.0 and 4.4

runif(15, -2.0, 4.4)

Optional tasks:

• Raise an error when the input file does not contain a header, or contains non-number items, or is otherwise inconsistent (not M x N numbers)
• Write a test: ./testGenRandMxes.sh , it should call another R script that gets two input paths (the original matrix and the directory with the noisified matrices) and a p value, and checks if the random noise falls within an acceptable range
• Turn the existing R codes into an R package and use testthat for testing