manifold_reduction: Compute Optimal Manifold Representation of Points

Description Usage Arguments Value References

View source: R/manifold.R

Description

This implements the algorithm described in Optimal Manifold Representation of Data: An Information Theoretic Approach by Denis Chigirev and William Bialek, which attempts to reduce higher dimensional data onto a lower dimensional manifold (by default 1D).

Usage

1
2
3
manifold_reduction(coords, lambda = 2, neighbourhood_size = 20L,
  knntouse = 75L, no_iterations = 45L, maxDim = 1.2,
  Verbose = TRUE, solvemethod = 0L)

Arguments

coords

d x N matrix of d-dimensional data points

lambda

A spatial parameter that determines the interactions between neighbouring points according to the tradeoff F(M,Pm) = D + lambda*I (see original paper for details).

neighbourhood_size

Number of nearest neighbours to consider when calculating local dimensionality (default 20).

knntouse

Number of nearest neighbours to consider when calculating interactions.

no_iterations

The number of iterations to use

maxDim

The local dimensionality below which points are fixed. The default value of 1.2 enforces a quasi-1D manifold.

Verbose

Whether to print status messages

solvemethod

An integer from -1 to 5 determining the solver used to compute the local dimensionality.

Value

a list with the following elements:

References

Optimal Manifold Representation of Data: An Information Theoretic Approach


jefferis/manifoldreduction documentation built on Oct. 25, 2019, 8:03 a.m.