Graph Interface

The Graph Interface provides functionality for creating graph-based representations of physics events.

Overview

Graphs are a powerful way to represent the relationships between particles in an event. The GraphTemplate class supports:

Node-based particle representation
Edge connectivity between particles
Feature extraction for nodes and edges
Integration with Graph Neural Networks

Core GraphTemplate Class

File Location

Cython Implementation: src/AnalysisG/core/graph_template.pyx
Cython Header: src/AnalysisG/core/graph_template.pxd

Class Definition

class GraphTemplate: Base class for graph representations of physics events.

Properties

property Nodes

List of graph nodes (typically particles).

Type:: list

property Edges

List of edges connecting nodes.

Type:: list[tuple]

property NodeFeatures

Feature matrix for nodes.

Type:: tensor

property EdgeFeatures

Feature matrix for edges.

Type:: tensor

Methods to Override

build_graph()

Construct the graph structure.

Override this method to define how particles are connected in the graph.

Usage Example

from AnalysisG.core import GraphTemplate

class MyGraph(GraphTemplate):
    def build_graph(self):
        # Add all particles as nodes
        for particle in self.event.Particles:
            self.add_node(particle)

        # Connect particles based on ΔR
        for i, p1 in enumerate(self.Nodes):
            for j, p2 in enumerate(self.Nodes):
                if i < j and p1.DeltaR(p2) < 0.4:
                    self.add_edge(i, j)