gng: Growing Neural Gas

Description Usage Arguments References

Description

Growing Neural Gas

Usage

1
2
3
4
gng(x, max_iter = 20000, epsilon_b = 0.05, epsilon_n = 0.001,
  age_max = 200, max_nodes = 30, lambda = 200, alpha = 0.5,
  beta = 0.99, assign_cluster = TRUE, verbose = TRUE, cpp = TRUE,
  make_logs_at = NULL)

Arguments

x

The input data. Must be a matrix!

max_iter

The max number of iterations.

epsilon_b

Move the winning node by epsilon_b times the distance.

epsilon_n

Move the neighbours of the winning node by epsilon_n times the distance.

age_max

Remove edges older than age_max.

max_nodes

The maximum number of nodes.

lambda

Insert new nodes every lambda iterations.

alpha

The decay parameter for error when a node is added.

beta

The decay parameter for error in every node every iteration.

assign_cluster

Whether or not to assign each sample to a GNG node.

verbose

Will output progress if TRUE.

cpp

Whether or not to use the C++ implementation over the R implementation. The C++ implementation is a lot faster.

make_logs_at

At which iterations to store the GNG, for visualisation purposes.

References

Fritzke, Bernd. "A growing neural gas network learns topologies." Advances in neural information processing systems 7 (1995): 625-632.


rcannood/gng documentation built on May 31, 2019, 7:54 a.m.