LFRContextBuilder

Overview

LFRContextBuilder is a ContextBuilder implementation for Repast Simphony that constructs agent-based simulation contexts using the LFR (Lancichinetti-Fortunato-Radicchi) benchmark network model. It handles agent creation, network generation, community assignment, and initial state configuration.

Key Functions

• Creates Agent instances and adds them to the simulation context
• Generates LFR networks via LFRNetworkGenerator with configurable parameters
• Maps network communities to individual agents
• Initializes agent properties (opinions, stubbornness, infection states) with community-based logic
• Supports network validation, statistics printing, and export (GML/CSV formats)

Core Methods

• build(Context<Object> context): Main entry point orchestrating context setup
• assignCommunityToAgents(): Maps LFR-generated communities to agents
• initializeAgentsWithStubbornCommunities(): Sets agent opinions/states (falls back to initializeAgentsOriginal() if insufficient communities)
• exportNetworkFiles(): Exports network data with timestamped filenames
• validateNetwork(): Checks network integrity (nodes, edges, isolated nodes, community distribution)

LFRNetworkGenerator

Overview

LFRNetworkGenerator<T> is a class implementing the LFR benchmark model, designed to generate synthetic networks with realistic community structures. This generator creates networks with power-law degree distributions and community size distributions, controlled by a mixing parameter (mu) that regulates the fraction of inter-community edges.

Key Features

• Generates networks with power-law degree distribution (controlled by tau1)
• Creates communities with power-law size distribution (controlled by tau2)
• Configurable mixing parameter (mu) for inter-community edge ratio
• Supports custom ranges for node degrees and community sizes
• Compatible with Repast Simphony's Network structure
• Maintains distribution properties while adjusting to target average degree

Constructor Parameters

Parameter	Type	Description
tau1	double	Power-law exponent for degree distribution (must be > 1)
tau2	double	Power-law exponent for community size distribution (must be > 1)
mu	double	Mixing parameter (0-1) controlling fraction of inter-community edges
minDegree	int	Minimum node degree (≥1)
maxDegree	int	Maximum node degree (≥ minDegree)
avgDegree	double	Target average degree (between minDegree and maxDegree)
minCommunity	int	Minimum community size
maxCommunity	int	Maximum community size (≥ minCommunity)
isSymmetrical	boolean	Whether the network is symmetrical

Core Methods

• createNetwork(Network<T> network): Main entry point that generates the complete LFR network, including degree sequences, communities, node assignments, and edges.
• generateDegreeSequence(int numNodes, boolean isDirected): Generates power-law distributed node degrees with gentle adjustment to match target average degree.
• generateCommunities(int numNodes): Creates community size distribution following power-law characteristics.
• assignNodesToCommunities(int numNodes): Distributes nodes to communities while maintaining power-law size properties.
• createEdges(Network<T> network, BidiMap<T, Integer> nodeMap): Builds edges with intra-community (1-μ) and inter-community (μ) fractions based on node degrees.

Export & Analysis

• exportToGML(Network<T> network, BidiMap<T, Integer> nodeMap, String filename): Exports network to GML format with node communities and edge types.
• exportToCSV(Network<T> network, BidiMap<T, Integer> nodeMap, String nodesFilename, String edgesFilename): Exports node/edge data to CSV for statistical analysis.
• calculateModularity(Network<T> network, BidiMap<T, Integer> nodeMap): Computes network modularity to evaluate community structure quality.

ScaleFreeNetwork

Overview

ScaleFreeNetwork is a class designed to generate scale-free networks using the Barabási-Albert (BA) model, tailored for agent-based simulations in Repast Simphony. Scale-free networks exhibit power-law degree distributions, where most nodes have few connections while a small number of hubs have many—mirroring real-world systems.

Key Features

• Implements the Barabási-Albert algorithm with preferential attachment
• Generates networks with power-law degree distributions
• Configurable via target average degree parameter
• Automatically adjusts for small networks (falls back to complete graphs)
• Calculates and reports network statistics (edge count, actual average degree)
• Supports opinion leader selection based on node degrees

Core Methods

• buildScaleFreeNetwork(Network<Object> network, List<Agent> agents, double averageDegree): Constructs the scale-free network by:

  • Creating an initial fully connected subgraph (m0 nodes)
  • Iteratively adding nodes with preferential attachment (m edges per new node)
  • Adjusting parameters for small agent populations

• calculateEdgeCount(Network<Object> network, List<Agent> agents): Computes total edge count (correcting for double-counting in undirected networks)

• selectByProbability(List<Agent> agents, List<Double> probabilities, Random random, List<Agent> excludeList): Implements preferential attachment by selecting nodes based on degree-weighted probabilities

• selectOpinionLeaders(Network<Object> network, List<Agent> agents, double opRatio): Identifies top-degree nodes as opinion leaders based on specified ratio

BA Model Parameters

• Derived from target averageDegree:
  • m: Number of edges per new node (≈ averageDegree / 2)
  • m0: Initial connected subgraph size (≥ m+1, minimum 3)

Usage Notes

• Automatically creates complete graphs for small agent counts (<= m0)
• Ensures no duplicate edges during network construction
• Includes fallback random selection if preferential attachment fails
• Prints network statistics (target/actual average degree, node/edge counts)