Interface for segment-edge GNN classification. More...

#include <ISegmentEdgeClassifierTool.h>

Inheritance diagram for MuonML::ISegmentEdgeClassifierTool:

Collaboration diagram for MuonML::ISegmentEdgeClassifierTool:

Public Member Functions
virtual	~ISegmentEdgeClassifierTool ()=default
	DeclareInterfaceID (ISegmentEdgeClassifierTool, 1, 0)
virtual StatusCode	buildGraph (const EventContext &ctx, const xAOD::MuonSegmentContainer &segments, SegmentEdgeGraph &graph) const =0
	Build a GNN graph from segments, extracting node features and computing edge connections based on kinematic compatibility.
virtual StatusCode	classifyEdges (const EventContext &ctx, const SegmentEdgeGraph &graph, std::vector< SegmentEdgeScore > &scores) const =0
	Run ONNX inference on the graph and return edge classification scores.

Detailed Description

Interface for segment-edge GNN classification.

This tool builds a graph representation of muon segments and classifies edges (potential segment pairs) using a graph neural network. Segments become nodes; edges connect nearby segments with compatible kinematics.

Workflow:

buildGraph() — Constructs node features from segment position/direction and bucket metadata, then builds sparse edges based on sector proximity and angular compatibility.
classifyEdges() — Runs ONNX inference on the graph to score each edge as a potential track connection and returns edge predictions.

Data Flow:

Input: xAOD::MuonSegmentContainer (raw segment collection)
Intermediate: SegmentEdgeGraph (nodes with features, sparse edge indices/features)
Output: std::vector<SegmentEdgeScore> (edge probabilities)

Definition at line 34 of file ISegmentEdgeClassifierTool.h.

Constructor & Destructor Documentation

◆ ~ISegmentEdgeClassifierTool()

virtual MuonML::ISegmentEdgeClassifierTool::~ISegmentEdgeClassifierTool ( )

virtualdefault

Member Function Documentation

◆ buildGraph()

virtual StatusCode MuonML::ISegmentEdgeClassifierTool::buildGraph	(	const EventContext &	ctx,
		const xAOD::MuonSegmentContainer &	segments,
		SegmentEdgeGraph &	graph ) const

pure virtual

Build a GNN graph from segments, extracting node features and computing edge connections based on kinematic compatibility.

Writes the graph structure (nodes, edges, and features) to graph.

Implemented in MuonML::SegmentEdgeClassifierTool.

◆ classifyEdges()

virtual StatusCode MuonML::ISegmentEdgeClassifierTool::classifyEdges	(	const EventContext &	ctx,
		const SegmentEdgeGraph &	graph,
		std::vector< SegmentEdgeScore > &	scores ) const

pure virtual

Run ONNX inference on the graph and return edge classification scores.

Reads the prepared graph and populates scores with logit and probability for each edge; the scores indicate the likelihood of a valid track connection between the two endpoint segments.