forward_mds: Do forward MDS by numerically minimizing the stress defined...

In NathanWycoff/mds.methods:

Description

Do forward MDS by numerically minimizing the stress defined in forward_cost.

Usage

forward_mds(high_d, weights, dist.func, thresh = 1e-05, max.iters = 1000,
  n.inits = 10, seed = NULL, std = TRUE, symm = FALSE)

Arguments

high_d	The high dimensional data of which a low dimensional representation is desired, an n by p matrix where rows represent observations
dist.func	The distance function to be used for high D distance computation; euclidean is always used for low D distance.
thresh	The threshold below which stress must fall before computation ends.
max.iters	The maximum number of iterations passed to optim for stress minimization.
n.inits	The number of times the optim is run from a random configuration. This can be important, as the cost surface is highly nonconvex.
seed	Random seed used for initialization
std	Boolean, should stress be standardized? This is accomplished by dividing stress by the sum of squared high D distance
symm	Boolean, is the distance function symmetric? We can save on computation if so.
low_d	The low_d solution, an n by 2 matrix, the cost of which is to be evaluated.

Value

List with $par, the optimal configuration as an n by 2 matrix, and $value as the stress of this configuration

NathanWycoff/mds.methods documentation built on May 23, 2019, 7:32 a.m.