Generate a Graphviz DOT Diagram for a Symmetric NMF Network

Creates a Graphviz DOT script that visualizes the two-layer structure of a symmetric NMF model (\(Y \approx X C X^\top\)).

The resulting diagram displays:

outer nodes: the original network nodes (rows/columns of Y),
inner nodes: the latent basis/group nodes (columns of X),
directed edges from basis to node: membership weights from X,
undirected edges between basis nodes: inter-group interactions from C matrix (for tri-symmetric models only).

Usage

nmfkc.net.DOT(
  result,
  threshold = 0.01,
  sig.level = 0.1,
  weight_scale = 5,
  weight_scale_xy = 1,
  weight_scale_xx = weight_scale,
  rankdir = "TB",
  fill = TRUE,
  hide.isolated = TRUE,
  Y.label = NULL,
  X.label = NULL,
  Y.title = "Nodes (Y)",
  X.title = "Basis (X)",
  show.theta = NULL,
  signed = inherits(result, "nmfkc.net.signed"),
  cluster.box = c("none", "normal", "faint", "invisible"),
  layout = c("fdp", "dot", "neato", "circo", "twopi"),
  X.color = NULL,
  Y.cluster = c("soft", "hard")
)

Arguments

result: A list returned by nmfkc() with Y.symmetric = "bi" or "tri", or the newer nmfkc.net with type = "bi" or "tri". If inference results are present (from nmfkc.net.inference), C edges are decorated with significance stars.
threshold: Minimum coefficient value to display an edge.
sig.level: Significance level for filtering C edges (if inference results are present). Set to NULL to show all edges above threshold.
weight_scale: Base scaling factor for edge widths.
weight_scale_xy: Scaling factor for X edges (basis -> node).
weight_scale_xx: Scaling factor for C edges (basis <-> basis).
rankdir: Graphviz rank direction ("TB" or "LR").
fill: Logical; whether nodes are filled with color.
hide.isolated: Logical; if TRUE, omit outer nodes with no X edge above threshold.
Y.label: Character vector of labels for outer nodes.
X.label: Character vector of labels for basis nodes.
Y.title: Cluster title for outer nodes.
X.title: Cluster title for basis nodes.
show.theta: Logical or NULL. Whether to draw C edges between basis nodes. NULL = auto-detect (TRUE for tri, FALSE for bi).
signed: Logical. If TRUE, \(C\) is treated as a signed matrix: positive entries rendered as solid edges and negative as dashed, with edge visibility threshold on \(|C|\). Default is inherits(result, "nmfkc.net.signed") so that results from nmfkc.net are auto-detected.
cluster.box: Style of cluster box: "none", "normal", "faint", "invisible".
layout: Graphviz layout engine: "fdp", "dot", "neato", "circo", "twopi".
X.color: Color palette for basis nodes (length Q).
Y.cluster: Coloring mode for outer nodes: "soft" (weighted mix) or "hard" (most probable basis color).

Value

A character string of class c("nmfkc.net.DOT", "nmfkc.DOT") representing a valid Graphviz DOT script. Use plot() to render (requires DiagrammeR).

Lifecycle

This function is experimental. The interface may change in future versions; details are to be described in an upcoming paper.

Examples

# \donttest{
library(nmfkc)
Y <- matrix(c(0,1,1,0,0,0,
              1,0,1,0,0,0,
              1,1,0,1,0,0,
              0,0,1,0,1,1,
              0,0,0,1,0,1,
              0,0,0,1,1,0), 6, 6)
res <- nmfkc.net(Y, rank = 2, type = "tri", nstart = 20)
dot <- nmfkc.net.DOT(res)
plot(dot)
# }