A small set of toy problems for anomaly detection and classification
One class problems: onemoon, thingaussian, and quadraticgaussian. All functions returns just a maxtrix with samples.
x = onemoon(n, sigma = 0.6,radius = 2) returns one moon from the famous two-moon problems
x = thinnormal(n, ϵ = 1e-1) returns a thin Gaussian (thinness controlled by ϵ) along the minor diagonal
x = quadraticnormal(n, ϵ = 1e-1) returns a one-dimensional Gaussian mapped to a parabola. ϵ controls the variance of additive noise
Two class problems: spirals and twomoon
(x,y) = spirals(n, k = 3)
(x,y) = twomoon(n, sigma = 0.6,radius = 2)