|
| 1 | +""" |
| 2 | +Graph Centrality Algorithms for Determining Central and Median Nodes in a Graph. |
| 3 | +
|
| 4 | +This module provides functions to compute the central and median nodes in a weighted |
| 5 | +graph based on graph-theoretical centrality measures. The central node minimizes the |
| 6 | +maximum shortest-path distance to all other reachable nodes (eccentricity), while the |
| 7 | +median node maximizes the sum of the reciprocals of the shortest-path distances to all |
| 8 | +other reachable nodes (harmonic closeness centrality). |
| 9 | +
|
| 10 | +Problem Description: |
| 11 | +Given a weighted graph G = (V, E), where V is the set of vertices, and E is the set of |
| 12 | +edges with positive weights representing distances between nodes, determine: |
| 13 | +
|
| 14 | +- Central Node: The node with minimal eccentricity. The eccentricity of a node v is |
| 15 | + defined as the greatest distance between v and any other node reachable from v. |
| 16 | +
|
| 17 | +- Median Node: The node with maximal harmonic closeness centrality. The harmonic |
| 18 | + closeness centrality of a node v is the sum of the reciprocals of the shortest-path |
| 19 | + distances from v to all other reachable nodes. |
| 20 | +
|
| 21 | +Algorithms Implemented: |
| 22 | +- Floyd-Warshall Algorithm for All-Pairs Shortest Paths. |
| 23 | +- Calculation of Eccentricity and Harmonic Closeness Centrality. |
| 24 | +
|
| 25 | +Algorithm Descriptions: |
| 26 | +
|
| 27 | +Floyd-Warshall Algorithm (Pseudo-code): |
| 28 | +--------------------------------------- |
| 29 | +for k from 1 to N: |
| 30 | + for i from 1 to N: |
| 31 | + for j from 1 to N: |
| 32 | + if distance[i][j] > distance[i][k] + distance[k][j]: |
| 33 | + distance[i][j] = distance[i][k] + distance[k][j] |
| 34 | +
|
| 35 | +Central and Median Node Calculation: |
| 36 | +------------------------------------ |
| 37 | +For each node i: |
| 38 | + - Eccentricity[i] = maximum distance from node i to any other reachable node. |
| 39 | + - Closeness[i] = sum of reciprocals of distances from node i to all reachable nodes. |
| 40 | +
|
| 41 | +Select: |
| 42 | + - Central Node: node with minimal eccentricity. |
| 43 | + - Median Node: node with maximal closeness. |
| 44 | +
|
| 45 | +References: |
| 46 | +- https://en.wikipedia.org/wiki/Centrality |
| 47 | +- Floyd-Warshall Algorithm: https://en.wikipedia.org/wiki/Floyd%E2%80%93Warshall_algorithm |
| 48 | +- Closeness Centrality: https://en.wikipedia.org/wiki/Closeness_centrality |
| 49 | +
|
| 50 | +Example Application: |
| 51 | +These algorithms can be applied to real-world problems, such as determining the optimal |
| 52 | +location for facilities (e.g., emergency response centers) to minimize response times |
| 53 | +within a network. By identifying the central or median nodes, organizations can make |
| 54 | +informed decisions on resource placement to improve efficiency and accessibility. |
| 55 | +""" |
| 56 | + |
| 57 | +import numpy as np |
| 58 | + |
| 59 | + |
| 60 | +def initialize_distance_matrix( |
| 61 | + graph: dict[int, list[tuple[int, float]]], number_of_nodes: int |
| 62 | +) -> np.ndarray: |
| 63 | + """Initialize the distance matrix and validate edge weights. |
| 64 | +
|
| 65 | + Args: |
| 66 | + graph: The graph represented as an adjacency list. |
| 67 | + number_of_nodes: The total number of nodes in the graph. |
| 68 | +
|
| 69 | + Returns: |
| 70 | + A numpy.ndarray representing the initialized distance matrix. |
| 71 | +
|
| 72 | + Raises: |
| 73 | + ValueError: If any edge has a non-positive weight. |
| 74 | + """ |
| 75 | + distance_matrix: np.ndarray = np.full((number_of_nodes, number_of_nodes), np.inf) |
| 76 | + np.fill_diagonal(distance_matrix, 0) |
| 77 | + |
| 78 | + for node_index, edges in graph.items(): |
| 79 | + for neighbor_index, edge_weight in edges: |
| 80 | + if edge_weight <= 0: |
| 81 | + error_message: str = ( |
| 82 | + f"Edge weight must be positive. Found {edge_weight} between " |
| 83 | + f"nodes {node_index} and {neighbor_index}." |
| 84 | + ) |
| 85 | + raise ValueError(error_message) |
| 86 | + distance_matrix[node_index, neighbor_index] = edge_weight |
| 87 | + |
| 88 | + return distance_matrix |
| 89 | + |
| 90 | + |
| 91 | +def floyd_warshall_algorithm(graph: dict[int, list[tuple[int, float]]]) -> np.ndarray: |
| 92 | + """Compute all-pairs shortest paths using the Floyd-Warshall algorithm. |
| 93 | +
|
| 94 | + Floyd-Warshall Complexity: |
| 95 | + -------------------------- |
| 96 | + Time Complexity: O(N^3), where N is the number of nodes. |
| 97 | + Space Complexity: O(N^2), for storing the distance matrix. |
| 98 | +
|
| 99 | + Args: |
| 100 | + graph: The graph represented as an adjacency list. |
| 101 | +
|
| 102 | + Returns: |
| 103 | + The distance matrix with the shortest paths between all pairs of nodes. |
| 104 | + """ |
| 105 | + number_of_nodes: int = len(graph) |
| 106 | + distance_matrix: np.ndarray = initialize_distance_matrix(graph, number_of_nodes) |
| 107 | + |
| 108 | + for k in range(number_of_nodes): |
| 109 | + # Use broadcasting to update the distance matrix in place |
| 110 | + distance_matrix[:] = np.minimum( |
| 111 | + distance_matrix, |
| 112 | + distance_matrix[:, k][:, np.newaxis] + distance_matrix[k, :], |
| 113 | + ) |
| 114 | + |
| 115 | + return distance_matrix |
| 116 | + |
| 117 | + |
| 118 | +def get_reachable_distances(distances: np.ndarray) -> np.ndarray: |
| 119 | + """Filter reachable distances, excluding infinite values (unreachable nodes). |
| 120 | +
|
| 121 | + Args: |
| 122 | + distances: Array of shortest-path distances from a specific node. |
| 123 | +
|
| 124 | + Returns: |
| 125 | + An array of distances to reachable nodes only (finite values). |
| 126 | + """ |
| 127 | + return distances[np.isfinite(distances) & (distances != 0)] |
| 128 | + |
| 129 | + |
| 130 | +def find_central_node( |
| 131 | + eccentricities: list[tuple[int, float]], |
| 132 | +) -> tuple[int, float]: |
| 133 | + """Identify the node with minimal eccentricity among reachable nodes. |
| 134 | +
|
| 135 | + Args: |
| 136 | + eccentricities: List of tuples (node index, eccentricity). |
| 137 | +
|
| 138 | + Returns: |
| 139 | + The node with minimal eccentricity and its value. Returns (-1, inf) if |
| 140 | + no valid nodes are found. |
| 141 | + """ |
| 142 | + valid_eccentricities = [e for e in eccentricities if e[1] != float("inf")] |
| 143 | + return min( |
| 144 | + valid_eccentricities, |
| 145 | + key=lambda node_eccentricity: (node_eccentricity[1], node_eccentricity[0]), |
| 146 | + default=(-1, float("inf")), |
| 147 | + ) |
| 148 | + |
| 149 | + |
| 150 | +def find_median_node(closenesses: list[tuple[int, float]]) -> tuple[int, float]: |
| 151 | + """Identify the node with maximal closeness among reachable nodes. |
| 152 | +
|
| 153 | + Args: |
| 154 | + closenesses: List of tuples (node index, closeness centrality). |
| 155 | +
|
| 156 | + Returns: |
| 157 | + The node with maximal closeness and its value. Returns (-1, inf) if |
| 158 | + no valid nodes are found. |
| 159 | + """ |
| 160 | + valid_closenesses = [c for c in closenesses if c[1] != float("inf")] |
| 161 | + return max( |
| 162 | + valid_closenesses, |
| 163 | + key=lambda node_closeness: (node_closeness[1], -node_closeness[0]), |
| 164 | + default=(-1, float("inf")), |
| 165 | + ) |
| 166 | + |
| 167 | + |
| 168 | +def find_central_and_median_node( |
| 169 | + distance_matrix: np.ndarray, |
| 170 | +) -> tuple[tuple[int, float], tuple[int, float]]: |
| 171 | + """Determine the central and median nodes based on shortest-path distances. |
| 172 | +
|
| 173 | + For each node, calculates its eccentricity and harmonic closeness centrality, |
| 174 | + considering only reachable nodes. Then, identifies the central node (minimal |
| 175 | + eccentricity) and median node (maximal closeness). |
| 176 | +
|
| 177 | + Args: |
| 178 | + distance_matrix: A numpy.ndarray representing shortest-path distances |
| 179 | + between all pairs of nodes. |
| 180 | +
|
| 181 | + Returns: |
| 182 | + A tuple containing: |
| 183 | + - central_node: A tuple (node index, eccentricity) for the node with |
| 184 | + minimal eccentricity. |
| 185 | + - median_node: A tuple (node index, closeness) for the node with maximal |
| 186 | + harmonic closeness centrality. |
| 187 | + """ |
| 188 | + num_nodes: int = len(distance_matrix) |
| 189 | + |
| 190 | + # Single-node graph case |
| 191 | + if num_nodes == 1: |
| 192 | + return (0, 0.0), (0, 0.0) |
| 193 | + |
| 194 | + eccentricities: list[tuple[int, float]] = [] |
| 195 | + closenesses: list[tuple[int, float]] = [] |
| 196 | + |
| 197 | + for i in range(num_nodes): |
| 198 | + reachable_distances = get_reachable_distances(distance_matrix[i]) |
| 199 | + |
| 200 | + if reachable_distances.size == 0: |
| 201 | + # No reachable nodes, isolated component |
| 202 | + eccentricity = float("inf") |
| 203 | + closeness = float("inf") |
| 204 | + else: |
| 205 | + # Compute eccentricity and closeness for reachable nodes |
| 206 | + eccentricity = float(np.max(reachable_distances)) |
| 207 | + closeness = float(np.sum(1 / reachable_distances)) |
| 208 | + |
| 209 | + eccentricities.append((i, eccentricity)) |
| 210 | + closenesses.append((i, closeness)) |
| 211 | + |
| 212 | + central_node = find_central_node(eccentricities) |
| 213 | + median_node = find_median_node(closenesses) |
| 214 | + |
| 215 | + return central_node, median_node |
| 216 | + |
| 217 | + |
| 218 | +# Test cases included as doctests |
| 219 | +def test_single_node() -> None: |
| 220 | + """ |
| 221 | + Test a graph with a single node. |
| 222 | +
|
| 223 | + >>> graph = {0: []} |
| 224 | + >>> distance_matrix = floyd_warshall_algorithm(graph) |
| 225 | + >>> central_node, median_node = find_central_and_median_node(distance_matrix) |
| 226 | + >>> central_node |
| 227 | + (0, 0.0) |
| 228 | + >>> median_node |
| 229 | + (0, 0.0) |
| 230 | + """ |
| 231 | + |
| 232 | + |
| 233 | +def test_two_nodes_positive_weight() -> None: |
| 234 | + """ |
| 235 | + Test a graph with two nodes connected by a positive weight. |
| 236 | +
|
| 237 | + >>> graph = {0: [(1, 5.0)], 1: [(0, 5.0)]} |
| 238 | + >>> distance_matrix = floyd_warshall_algorithm(graph) |
| 239 | + >>> central_node, median_node = find_central_and_median_node(distance_matrix) |
| 240 | + >>> central_node |
| 241 | + (0, 5.0) |
| 242 | + >>> median_node |
| 243 | + (0, 0.2) |
| 244 | + """ |
| 245 | + |
| 246 | + |
| 247 | +def test_fully_connected_graph() -> None: |
| 248 | + """ |
| 249 | + Test a fully connected graph. |
| 250 | +
|
| 251 | + >>> graph = { |
| 252 | + ... 0: [(1, 1.0), (2, 1.0)], |
| 253 | + ... 1: [(0, 1.0), (2, 1.0)], |
| 254 | + ... 2: [(0, 1.0), (1, 1.0)], |
| 255 | + ... } |
| 256 | + >>> distance_matrix = floyd_warshall_algorithm(graph) |
| 257 | + >>> central_node, median_node = find_central_and_median_node(distance_matrix) |
| 258 | + >>> central_node |
| 259 | + (0, 1.0) |
| 260 | + >>> median_node |
| 261 | + (0, 2.0) |
| 262 | + """ |
| 263 | + |
| 264 | + |
| 265 | +def test_directed_acyclic_graph() -> None: |
| 266 | + """ |
| 267 | + Test a directed acyclic graph (DAG). |
| 268 | +
|
| 269 | + >>> graph = { |
| 270 | + ... 0: [(1, 1.0), (2, 2.0)], |
| 271 | + ... 1: [(3, 3.0)], |
| 272 | + ... 2: [(3, 1.0)], |
| 273 | + ... 3: [] |
| 274 | + ... } |
| 275 | + >>> distance_matrix = floyd_warshall_algorithm(graph) |
| 276 | + >>> central_node, median_node = find_central_and_median_node(distance_matrix) |
| 277 | + >>> central_node |
| 278 | + (2, 1.0) |
| 279 | + >>> median_node |
| 280 | + (0, 1.8333333333333333) |
| 281 | + """ |
| 282 | + |
| 283 | + |
| 284 | +def test_disconnected_graph() -> None: |
| 285 | + """ |
| 286 | + Test a disconnected graph with nodes that cannot reach each other. |
| 287 | +
|
| 288 | + >>> graph = { |
| 289 | + ... 0: [], |
| 290 | + ... 1: [], |
| 291 | + ... 2: [] |
| 292 | + ... } |
| 293 | + >>> distance_matrix = floyd_warshall_algorithm(graph) |
| 294 | + >>> central_node, median_node = find_central_and_median_node(distance_matrix) |
| 295 | + >>> central_node |
| 296 | + (-1, inf) |
| 297 | + >>> median_node |
| 298 | + (-1, inf) |
| 299 | + """ |
| 300 | + |
| 301 | + |
| 302 | +def test_graph_with_zero_weight() -> None: |
| 303 | + """ |
| 304 | + Test a graph with zero weight, which should raise a ValueError. |
| 305 | +
|
| 306 | + >>> graph = {0: [(1, 0.0)], 1: []} |
| 307 | + >>> floyd_warshall_algorithm(graph) |
| 308 | + Traceback (most recent call last): |
| 309 | + ... |
| 310 | + ValueError: Edge weight must be positive. Found 0.0 between nodes 0 and 1. |
| 311 | + """ |
| 312 | + |
| 313 | + |
| 314 | +def test_graph_with_negative_weight() -> None: |
| 315 | + """ |
| 316 | + Test a graph with negative weight, which should raise a ValueError. |
| 317 | +
|
| 318 | + >>> graph = {0: [(1, -2.0)], 1: []} |
| 319 | + >>> floyd_warshall_algorithm(graph) |
| 320 | + Traceback (most recent call last): |
| 321 | + ... |
| 322 | + ValueError: Edge weight must be positive. Found -2.0 between nodes 0 and 1. |
| 323 | + """ |
| 324 | + |
| 325 | + |
| 326 | +def test_cyclic_graph() -> None: |
| 327 | + """ |
| 328 | + Test a cyclic graph where there is a cycle between nodes. |
| 329 | +
|
| 330 | + >>> graph = { |
| 331 | + ... 0: [(1, 1.0)], |
| 332 | + ... 1: [(2, 1.0)], |
| 333 | + ... 2: [(0, 1.0)] |
| 334 | + ... } |
| 335 | + >>> distance_matrix = floyd_warshall_algorithm(graph) |
| 336 | + >>> central_node, median_node = find_central_and_median_node(distance_matrix) |
| 337 | + >>> central_node |
| 338 | + (0, 2.0) |
| 339 | + >>> median_node |
| 340 | + (0, 1.5) |
| 341 | + """ |
| 342 | + |
| 343 | + |
| 344 | +def test_sparse_graph() -> None: |
| 345 | + """ |
| 346 | + Test a larger sparse graph. |
| 347 | +
|
| 348 | + >>> graph = { |
| 349 | + ... 0: [(1, 2.0)], |
| 350 | + ... 1: [(2, 3.0)], |
| 351 | + ... 2: [(3, 4.0)], |
| 352 | + ... 3: [(4, 5.0)], |
| 353 | + ... 4: [] |
| 354 | + ... } |
| 355 | + >>> distance_matrix = floyd_warshall_algorithm(graph) |
| 356 | + >>> central_node, median_node = find_central_and_median_node(distance_matrix) |
| 357 | + >>> central_node |
| 358 | + (3, 5.0) |
| 359 | + >>> median_node |
| 360 | + (0, 0.8825396825396825) |
| 361 | + """ |
| 362 | + |
| 363 | + |
| 364 | +def test_large_fully_connected_graph() -> None: |
| 365 | + """ |
| 366 | + Test a larger fully connected graph with random weights. |
| 367 | +
|
| 368 | + >>> import random |
| 369 | + >>> random.seed(42) |
| 370 | + >>> number_of_nodes = 10 |
| 371 | + >>> graph = {i: [(j, random.uniform(1, 10)) for j in |
| 372 | + ... range(number_of_nodes) if i != j] |
| 373 | + ... for i in range(number_of_nodes)} |
| 374 | + >>> distance_matrix = floyd_warshall_algorithm(graph) |
| 375 | + >>> central_node, median_node = find_central_and_median_node(distance_matrix) |
| 376 | + >>> central_node[0] is not None # Ensure it found a central node |
| 377 | + True |
| 378 | + >>> median_node[0] is not None # Ensure it found a median node |
| 379 | + True |
| 380 | + """ |
| 381 | + |
| 382 | + |
| 383 | +if __name__ == "__main__": |
| 384 | + import doctest |
| 385 | + |
| 386 | + doctest.testmod() |
0 commit comments