dimfactors: Specifying SOD dimensionality

In SOD: SOD for multidimensional scaling.

Description

Functions that provide a matrix of dimensionality factors that can be used by the R_DimSqueezer squeezeDF() function.
During the mapping process the dimensionality is gradually reduced. The dimensionality at a given time is represented by a vector of values (one for each dimension) lying within the range 0-1.
0 indicates a null dimension; i.e. coordinates within that dimension have no impact upon distances between objects. A value of 1 indicates a full dimension.
These functions return matrices that define the dimensionality with one row for each iteration of the mapping process.

• parallelDimFactors,
  parallelExpDimFactors
  Reduce the superfluous dimensions in parallel, using a linear or exponential decrease respectively.

• serialDimFactors
  Reduces superfluous dimensions serially one after the other.

Usage

parallelDimFactors(dim, iteration.no,
                     red.end=iteration.no*0.75, target.dim=2)

parallelExpDimFactors(dim, iteration.no,
                         target.dim=2, red.end=iteration.no * 0.9)

serialDimFactors(dim, iteration.no,
                   red.end=iteration.no*0.75, target.dim=2)

Arguments

dim	The starting number of dimensions
target.dim	The target number of dimensions
iteration.no	The total number of iterations for the mapping procedure
red.end	The number of the iterations which should be used to remove superflous dimensions. Iterations after this are used to adjust the final positions.

Value

Returns a numeric matrix with iteration.no rows and dim columns.

Examples

## reduce to two dimensions using (0.75 * 200) dimensions
## 
df.p <- parallelDimFactors(6, 200, target.dim=2)

## start with 6 dimensions, and reduce over 175 iterations
## to 2 dimensions.
df.pe <- parallelExpDimFactors(6, 200, 175, target.dim=2)

SOD documentation built on May 30, 2017, 2:03 a.m.