principal_graph: function to automatically learn the structure of data by...
In cole-trapnell-lab/L1-graph: Finding Principal Graphs via L1 Optimization

function to automatically learn the structure of data by either using L1-graph or the spanning-tree formulization

1
2
3

principal_graph(X, C0, G, maxiter = 10, eps = 1e-05,
  gstruct = c("l1-graph", "span-tree"), L1.lambda = 1, L1.gamma = 0.5,
  L1.sigma = 0.01, nn = 5, L1.timeout = 1800, verbose = T)

`X`	the input data DxN
`C0`	the initialization of centroids
`G`	graph matrix with side information where cannot-link pair is 0
`maxiter`	maximum number of iteraction
`eps`	relative objective difference
`gstruct`	graph structure to learn, either L1-graph or the span-tree
`L1.lambda`	regularization parameter for inverse graph embedding
`L1.gamma`	regularization parameter for k-means (the prefix of 'param' is used to avoid name collision with gamma)
`L1.sigma`	bandwidth parameter
`nn`	number of nearest neighbors
`L1.timeout`	a positive integer value specifying the number of seconds after which a timeout will occur. If zero, then no timeout will occur. (This is a parameter passed to lp.control function)
`verbose`	emit results from iteraction

a list of X, C, W, P, objs X is the input data C is the centers for principal graph W is the pricipal graph matrix P is the cluster assignment matrix objs is the objective value for the function

cole-trapnell-lab/L1-graph documentation built on May 17, 2019, 12:50 p.m.