ReduceDimension: Reduce the dimensionality of the optimization

In google/amss: Agreggate Marketing System Simulator

Description

Reduce the dimensionality of the optimization and find a function that maps the new search space of the optimization to the original search space.

Usage

ReduceDimension(n.dim, lower.bound, upper.bound, sum.lower.bound,
  sum.upper.bound)

Arguments

n.dim	integer, original dimensionality of the optimization.
lower.bound	numeric vector, the constraint specifying the lower bound in each dimension.
upper.bound	numeric vector, the constraint specifying the upper bound in each dimension.
sum.lower.bound	numeric, the lower bound on the vector sum.
sum.upper.bound	numeric, the upper bound on the vector sum.

Details

The search space is assumed to be the intersection of a bounding box and an upper/lower bound on the vector sum. If possible, the function finds an equivalent lower-dimensional search space of the same type, and provides a function mapping the new search space to the old search space.

Value

ReduceDimension returns NULL if the constrained set is empty. Else, it returns a codelist with elements

lower.bound

the new lower bound.

upper.bound

the new upper bound.

sum.lower.bound

the new lower bound on the vector sum.

sum.upper.bound

the new upper bound on the vector sum.

decoder

function mapping values in the new search space to values in the original search space.

encoder

function mapping values in the original search space to values in the new search space.

google/amss documentation built on May 20, 2019, 5:05 p.m.