Examples

This section provides practical examples demonstrating various features of egraph.

• Stochastic Gradient Descent (SGD)
  • Basic SGD Example
  • Using FullSgd
• Stress Majorization
  • Basic Stress Majorization Example
  • Using a Distance Matrix
  • Applying a Single Iteration
• Kamada-Kawai
  • Basic Kamada-Kawai Example
  • Using Custom Edge Distances
  • Applying to a Single Node
• 3D Stochastic Gradient Descent
  • Basic 3D SGD Example
  • Using ClassicalMds for 3D Initialization
• Hyperbolic 2D Stochastic Gradient Descent
  • Basic Hyperbolic SGD Example
  • Working with Hyperbolic Distances
• Spherical 2D Stochastic Gradient Descent
  • Basic Spherical SGD Example
  • Working with Spherical Distances
• Torus 2D Stochastic Gradient Descent
  • Basic Torus SGD Example
  • Visualizing the Torus in 3D
• Overwrap Removal
  • Basic Overwrap Removal Example
  • Using Variable Node Sizes
  • Controlling the Overlap Removal Process

Overview

These examples demonstrate real-world usage of egraph’s features. Each example is self-contained and can be run independently.

Layout Algorithms

• Stochastic Gradient Descent (SGD) - Stochastic Gradient Descent for fast, scalable layouts

• Stress Majorization - High-quality layouts through stress minimization

• Kamada-Kawai - Spring-based layout algorithm

Advanced Drawing Spaces

• 3D Stochastic Gradient Descent - Three-dimensional graph layouts

• Hyperbolic 2D Stochastic Gradient Descent - Hyperbolic space for hierarchical graphs

• Spherical 2D Stochastic Gradient Descent - Spherical layouts for global networks

• Torus 2D Stochastic Gradient Descent - Torus layouts with periodic boundaries

Specialized Features

• Overwrap Removal - Eliminate node overlaps while preserving structure

Quick Example

Here’s a simple example of creating a graph and applying a layout algorithm:

import networkx as nx
import egraph as eg

# Create a graph from NetworkX
nx_graph = nx.les_miserables_graph()
graph = eg.Graph()
indices = {}
for u in nx_graph.nodes:
    indices[u] = graph.add_node(u)
for u, v in nx_graph.edges:
    graph.add_edge(indices[u], indices[v], (u, v))

# Create an initial drawing
drawing = eg.DrawingEuclidean2d.initial_placement(graph)

# Apply a layout algorithm
sm = eg.StressMajorization(graph, drawing, lambda _: 100)
sm.run(drawing)

# Extract node positions
pos = {u: (drawing.x(i), drawing.y(i)) for u, i in indices.items()}

# Visualize with NetworkX
import matplotlib.pyplot as plt
nx.draw(nx_graph, pos)