makeNotNotMnistProblem: MNIST like problem.
In ja-thomas/dlbench: Deep Learning Benchmark Problems

Description Usage Arguments Value Examples

View source: R/makeNotNotMnist.R

Generate images of letters or numbers in various ways. Randomized size, position, rotations and fonts are available to set the difficulty of the problem.

1
2
3

makeNotNotMnistProblem(n = 5^4, sample.font = TRUE, random.rotation = 0,
  position.range = c(0.3, 0.7), size.range = c(15, 50), letters = LETTERS,
  resolution = c(32, 32), pointsize = 2, create.task = TRUE)

`n`	[`integer(1)`] Number of observations to create. Default is `5000L`.
`sample.font`	[`logical(1)`] Should every image be sampled from a different font. Fonts are sampled from `Hershey`. Default is `TRUE`.
`random.rotation`	[`numeric(1)`] Should the rotation of the letters/numbers be randomized. Value is the maximum degree of rotation. Default is 0, which is no rotation at all.
`position.range`	[`numeric(2`] Randomize position of the letter/number. First number is the lower bound for x/y value, second number the upper bound. Minimal/maximal values of the canvas are 0 and 1. Note that it is possible that some element of the letters/number might be out of the bound, depending on the range and size. Default is `c(0.3, 0.7)`
`size.range`	[`numeric(2)`] Size of the letters/numbers, is sampled between first and second value. This is the `cex` value of the plot. Default is `c(15, 50)`.
`letters`	[`numeric() \| character()`] Letters/Numbers to sample from. Can be anything that can be passed to `text()`. Default is `LETTERS`.
`resolution`	[`integer(2)`] Resolution of the image. This defines the number of features. Default is `c(32, 32)`, which creates `1024` features.
`pointsize`	[`integer(1)`] Pointsize of intepreted big points (1/72 inch). See `bmp`. Default is `2`.
`create.task`	[`logical(1)`] Should a task be created? If `TRUE` a mlr classification task (see `makeClassifTask` is returned. Otherwise a `data.frame` is returned.