我有从S到T的有向图。
我想找到路线(S,A,C,E,T)及其容量之和(1 + 2 + 3 + 1 = 7),因此总和最大。
我尝试过networkx.algorithms.flow.ford_fulkerson,但我不知道如何从S到T获得单向指示。
我的环境:
example.py
#!/usr/bin/env python
# -*- coding: utf-8 -*-
import matplotlib.pylab as p
import networkx as nx
if __name__ == '__main__':
DG = nx.DiGraph()
DG.add_edge('S', 'a', capacity=1)
DG.add_edge('a', 'b', capacity=1)
DG.add_edge('a', 'c', capacity=2)
DG.add_edge('b', 'd', capacity=1)
DG.add_edge('b', 'e', capacity=2)
DG.add_edge('c', 'e', capacity=3)
DG.add_edge('c', 'f', capacity=2)
DG.add_edge('d', 'T', capacity=1)
DG.add_edge('e', 'T', capacity=1)
DG.add_edge('f', 'T', capacity=1)
result = nx.algorithms.flow.ford_fulkerson(DG, 'S', 'T')
print(result.size(weight='capacity')) # 15.0, but I want 7.
pos = nx.spectral_layout(DG)
nx.draw(DG, pos)
nx.draw_networkx_labels(DG, pos)
nx.draw_networkx_edge_labels(DG, pos)
p.show()
# This shows a partly bidirectional graph, which is not what I want.
pos = nx.spectral_layout(result)
nx.draw(result, pos)
nx.draw_networkx_labels(result, pos)
nx.draw_networkx_edge_labels(result, pos)
p.show()
答案 0 :(得分:1)
使用负权重通常不适用于Dijkstra算法。
此错误ValueError: ('Contradictory paths found:', 'negative weights?')
将会显示。
它应该区分“最长路径”和“最大总和路径”的问题。
答案在这里:How to find path with highest sum in a weighted networkx graph?,使用all_simple_paths。
请注意,在函数all_simple_paths(G, source, target, cutoff=None)
中,使用cutoff
param(整数)可以帮助限制从source
到target
的搜索深度。它还控制我们想要找到的路径的长度。
答案 1 :(得分:1)
负重量适用于约翰逊。在您的情况下,修改为:
DG = nx.DiGraph()
DG.add_edge('S', 'a', weight=-1)
DG.add_edge('a', 'b', weight=-1)
DG.add_edge('a', 'c', weight=-2)
DG.add_edge('b', 'd', weight=-1)
DG.add_edge('b', 'e', weight=-2)
DG.add_edge('c', 'e', weight=-3)
DG.add_edge('c', 'f', weight=-2)
DG.add_edge('d', 'T', weight=-1)
DG.add_edge('e', 'T', weight=-1)
DG.add_edge('f', 'T', weight=-1)
要找到最长的路径,请使用
获取从S到T的单向指示p2 = nx.johnson (DG, weight='weight')
print('johnson: {0}'.format(p2['S']['T']))
结果为:johnson: ['S', 'a', 'c', 'e', 'T']
我的环境:
答案 2 :(得分:0)
我想我找到了解决方案。
#!/usr/bin/env python
# -*- coding: utf-8 -*-
import networkx as nx
def inverse_weight(graph, weight='weight'):
copy_graph = graph.copy()
for n, eds in copy_graph.adjacency_iter():
for ed, eattr in eds.items():
copy_graph[n][ed][weight] = eattr[weight] * -1
return copy_graph
def longest_path_and_length(graph, s, t, weight='weight'):
i_w_graph = inverse_weight(graph, weight)
path = nx.dijkstra_path(i_w_graph, s, t)
length = nx.dijkstra_path_length(i_w_graph, s, t) * -1
return path, length
if __name__ == '__main__':
DG = nx.DiGraph()
DG.add_edge('S', 'a', weight=1)
DG.add_edge('a', 'b', weight=1)
DG.add_edge('a', 'c', weight=2)
DG.add_edge('b', 'd', weight=1)
DG.add_edge('b', 'e', weight=2)
DG.add_edge('c', 'e', weight=3)
DG.add_edge('c', 'f', weight=2)
DG.add_edge('d', 'T', weight=1)
DG.add_edge('e', 'T', weight=1)
DG.add_edge('f', 'T', weight=1)
print(longest_path_and_length(DG, 'S', 'T')) # (['S', 'a', 'c', 'e', 'T'], 7)