A matrix-based representation of a network or graph where rows and columns represent nodes and cell color indicates the presence or weight of edges between them. This visualization complements node-link diagrams by excelling at revealing clusters, structural patterns, and density in large or dense networks where node-link layouts become cluttered. Reordering nodes by cluster, degree, or community membership exposes block-diagonal structure and makes group boundaries immediately visible.
""" anyplot.ai
heatmap-adjacency: Network Adjacency Matrix Heatmap
Library: plotnine 0.15.4 | Python 3.13.13
Quality: 88/100 | Created: 2026-05-08
"""
import os
import numpy as np
import pandas as pd
from plotnine import (
aes,
element_blank,
element_rect,
element_text,
geom_hline,
geom_tile,
geom_vline,
ggplot,
labs,
scale_fill_cmap,
scale_x_discrete,
scale_y_discrete,
theme,
)
# Theme tokens
THEME = os.getenv("ANYPLOT_THEME", "light")
PAGE_BG = "#FAF8F1" if THEME == "light" else "#1A1A17"
ELEVATED_BG = "#FFFDF6" if THEME == "light" else "#242420"
INK = "#1A1A17" if THEME == "light" else "#F0EFE8"
INK_SOFT = "#4A4A44" if THEME == "light" else "#B8B7B0"
ABSENT_CELL = "#E0DED7" if THEME == "light" else "#2A2A25"
# Data: researcher collaboration network — 3 scientific fields (Physics, Biology, CompSci)
np.random.seed(42)
n_per_group = 10
prefixes = ["P", "B", "C"]
node_names = [f"{p}{i + 1:02d}" for p in prefixes for i in range(n_per_group)]
n_nodes = len(node_names)
W = np.zeros((n_nodes, n_nodes))
# Within-field: dense, strong collaborations (block-diagonal structure)
for gi in range(3):
members = list(range(gi * n_per_group, (gi + 1) * n_per_group))
for idx_i, i in enumerate(members):
for j in members[idx_i + 1 :]:
if np.random.rand() < 0.72:
w = np.random.uniform(0.45, 1.0)
W[i, j] = w
W[j, i] = w
# Cross-field: sparse, weaker connections
for i in range(n_nodes):
for j in range(i + 1, n_nodes):
if i // n_per_group != j // n_per_group and W[i, j] == 0:
if np.random.rand() < 0.09:
w = np.random.uniform(0.05, 0.38)
W[i, j] = w
W[j, i] = w
# Long-format DataFrame; absent edges (including diagonal) → NaN
all_pairs = [(r, c) for r in range(n_nodes) for c in range(n_nodes)]
df = pd.DataFrame(
{
"source": pd.Categorical([node_names[r] for r, _ in all_pairs], categories=node_names, ordered=True),
"target": pd.Categorical([node_names[c] for _, c in all_pairs], categories=node_names, ordered=True),
"weight": [W[r, c] if W[r, c] > 0 else np.nan for r, c in all_pairs],
}
)
# Community separator positions (between groups of 10)
boundaries = [n_per_group + 0.5, 2 * n_per_group + 0.5]
# Plot
anyplot_theme = theme(
figure_size=(12, 12),
plot_background=element_rect(fill=PAGE_BG, color=PAGE_BG),
panel_background=element_rect(fill=ABSENT_CELL),
panel_grid_major=element_blank(),
panel_grid_minor=element_blank(),
panel_border=element_blank(),
axis_title=element_text(color=INK, size=20),
axis_text=element_text(color=INK_SOFT, size=16),
axis_text_x=element_text(angle=45, ha="right", color=INK_SOFT, size=16),
plot_title=element_text(color=INK, size=24, ha="center"),
legend_background=element_rect(fill=ELEVATED_BG, color=None),
legend_text=element_text(color=INK_SOFT, size=16),
legend_title=element_text(color=INK, size=18),
)
plot = (
ggplot(df, aes(x="target", y="source", fill="weight"))
+ geom_tile()
+ geom_vline(xintercept=boundaries, color=INK, size=0.9, alpha=0.5, linetype="dashed")
+ geom_hline(yintercept=boundaries, color=INK, size=0.9, alpha=0.5, linetype="dashed")
+ scale_fill_cmap("viridis", na_value=ABSENT_CELL, name="Collaboration\nStrength")
+ scale_x_discrete(limits=node_names)
+ scale_y_discrete(limits=node_names[::-1])
+ labs(x="Target Researcher", y="Source Researcher", title="heatmap-adjacency · plotnine · anyplot.ai")
+ anyplot_theme
)
# Save
plot.save(f"plot-{THEME}.png", dpi=300, width=12, height=12, units="in")
Part of Network Adjacency Matrix Heatmap on anyplot.ai.