Updated data generation code

generate_data_consensus.py

@@ -8,7 +8,9 @@ import networkx as nx
 import networkx.algorithms.community as nx_comm
 from tqdm import tqdm
 import sims
+from dotenv import load_dotenv
 
+load_dotenv()
 
 
 def generate_connected_graph(rng: np.random.Generator, num_agents: int, graph_type: str) -> tuple[nx.Graph, str, np.ndarray]:
@@ -21,35 +23,34 @@ def generate_connected_graph(rng: np.random.Generator, num_agents: int, graph_ty
     """
 
     G = None
-    
 
     if graph_type == "erdos_renyi":
-    
-        # p = 2.0 * np.log(num_agents) / num_agents
         p = 0.5
         G = nx.erdos_renyi_graph(num_agents, p, seed=rng)
     
     elif graph_type == "watts_strogatz":
-        k = 2
-        p = 0.1
+        k = 4
+        p = 0.5
         G = nx.watts_strogatz_graph(num_agents, k, p, seed=rng)
     
     elif graph_type == "barabasi_albert":
-        m = 2
-        if m >= num_agents: m = max(1, num_agents - 1)
-        G = nx.barabasi_albert_graph(num_agents, m, seed=rng)
+        G = nx.barabasi_albert_graph(num_agents, 2, seed=rng)
     
     elif  graph_type == "powerlaw_cluster":
-        m = 3 # Number of random edges to add for each new node
-        p = 0.1 # Probability of adding a triangle after adding a random edge
-        if m >= num_agents: m = max(1, num_agents - 1)
+        m = 2 
+        p = 0.5 
         G = nx.powerlaw_cluster_graph(num_agents, m, p, seed=rng)
 
-    # Add self-loops, as they are often assumed in consensus algorithms
-    G.add_edges_from([(i, i) for i in range(num_agents)])
-    adj_matrix = nx.to_numpy_array(G, dtype=np.float32)
 
-    return G, graph_type, adj_matrix
+    graph_matrix = nx.adjacency_matrix(G).todense()
+
+    for i in range(num_agents):
+        graph_matrix[i,i]=1
+
+    G = nx.Graph(graph_matrix)
+
+    return G, graph_type, graph_matrix
+
 
 def calculate_graph_metrics(G: nx.Graph) -> dict:
     """
@@ -91,30 +92,30 @@ def calculate_graph_metrics(G: nx.Graph) -> dict:
 
     # --- Spectral & Community Metrics (Potentially Slow) ---
     # These are also limited to smaller graphs.
-    if 1 < num_nodes < 500:
-        # Eigenvalues of the Laplacian matrix
-        try:
-            laplacian_eigenvalues = sorted(nx.laplacian_spectrum(G))
-            metrics["laplacian_eigenvalues"] = laplacian_eigenvalues
-            # The second-smallest eigenvalue of the Laplacian matrix
-            metrics["algebraic_connectivity"] = laplacian_eigenvalues[1]
-        except Exception:
-            metrics["laplacian_eigenvalues"] = None
-            metrics["algebraic_connectivity"] = None
+    # if 1 < num_nodes < 500:
+    #     # Eigenvalues of the Laplacian matrix
+    #     try:
+    #         laplacian_eigenvalues = sorted(nx.laplacian_spectrum(G))
+    #         metrics["laplacian_eigenvalues"] = laplacian_eigenvalues
+    #         # The second-smallest eigenvalue of the Laplacian matrix
+    #         metrics["algebraic_connectivity"] = laplacian_eigenvalues[1]
+    #     except Exception:
+    #         metrics["laplacian_eigenvalues"] = None
+    #         metrics["algebraic_connectivity"] = None
 
-        # Modularity using the Louvain community detection algorithm
-        if num_edges > 0:
-            try:
-                communities = nx_comm.louvain_communities(G, seed=123)
-                metrics["modularity"] = nx_comm.modularity(G, communities)
-            except Exception:
-                metrics["modularity"] = None # Algorithm may fail on some graphs
-        else:
-            metrics["modularity"] = None
-    else:
-        metrics["laplacian_eigenvalues"] = None
-        metrics["algebraic_connectivity"] = None
-        metrics["modularity"] = None
+    #     # Modularity using the Louvain community detection algorithm
+    #     if num_edges > 0:
+    #         try:
+    #             communities = nx_comm.louvain_communities(G, seed=123)
+    #             metrics["modularity"] = nx_comm.modularity(G, communities)
+    #         except Exception:
+    #             metrics["modularity"] = None # Algorithm may fail on some graphs
+    #     else:
+    #         metrics["modularity"] = None
+    # else:
+    #     metrics["laplacian_eigenvalues"] = None
+    #     metrics["algebraic_connectivity"] = None
+    #     metrics["modularity"] = None
         
     return metrics
 
@@ -132,9 +133,9 @@ def main():
     # --- Configuration ---
     GRAPH_GEN_ALGOS = ["erdos_renyi", "barabasi_albert", "powerlaw_cluster", "watts_strogatz"]
     AGENT_COUNTS = range(5, 51, 5)# 5, 10, ..., 50 agents
-    GRAPHS_PER_AGENT_COUNT = 100
+    GRAPHS_PER_AGENT_COUNT = 80
     GRAPHS_PER_GRAPH_ALGO = GRAPHS_PER_AGENT_COUNT // len(GRAPH_GEN_ALGOS)
-    SIMS_PER_GRAPH = 100
+    SIMS_PER_GRAPH = 400
     OUTPUT_DIR = "datasets/consensus_dataset"
     
     # --- Setup ---