上下界费用流

CodeForces 708D Incorrect Flow

这是道可行流最小费用的题目,这题好神奇。

看这老哥的题解就ok了。戳这

我觉得他少说了一条边,但代码中却体现了,就是一条流量为$(f,f)$的边,费用为0,就是固定流量的一条边。

 还要注意一点是,$f>c$的时候,$f$往下调到$c$以下时,$c$是不动的。

#include <bits/stdc++.h>
using namespace std;
const int maxn = 200;
const int INF = 1e9;
int dist[maxn];
int pv[maxn],pe[maxn];
struct edge
{
    int to, cap, pre;
    int cost;
}e[1450];
int tot = 0, head[maxn];
void init(int n)
{
    tot = 0;
    fill(head, head + n + 1, -1);
}
void add(int from,int to,int cap, int cost)
{
    e[tot].pre = head[from];
    e[tot].to = to;
    e[tot].cap = cap;
    e[tot].cost = cost;
    head[from] = tot++;
}
void addedge(int from,int to,int cap, int cost)
{
    add(from,to,cap,cost);
    add(to,from,0,-cost);
}
int vis[maxn];
void SPFA(int s, int t)
{
    for(int i = 1; i < maxn; i++) dist[i] = INF;
    memset(vis, 0, sizeof(vis));
    dist[s] = 0, vis[s] = 1;
    queue<int> q;
    q.push(s);
    while(!q.empty())
    {
        int u = q.front();
        q.pop();
        vis[u] = 0;
        for(int i = head[u]; ~i; i = e[i].pre)
        {
            int to = e[i].to, cap = e[i].cap;
            if(cap > 0 && (dist[to] - (dist[u] + e[i].cost)) > 0)
            {
                pv[to] = u, pe[to] = i;
                dist[to] = dist[u] + e[i].cost;
                if(!vis[to])
                {
                    vis[to] = 1;
                    q.push(to);
                }
            }
        }
    }
}
int min_cost_flow(int s,int t,int f,int& max_flow)
{
    int ret = 0;
    while(f>0)
    {
        SPFA(s, t);
        if(dist[t] == INF) return ret;///同一目的地,每次增广路都是最小费用
        ///当所有边的流量都流净后,即没有残余网络,返回。
        int d = f;
        for(int v=t;v!=s;v=pv[v])
        {
            d = min(d, e[pe[v]].cap);
        }
        f -= d;
        max_flow += d;
        ret += (int)d*dist[t]; ///走一单位就消耗dist[t]
        for(int v=t;v!=s;v=pv[v])
        {
            e[pe[v]].cap -= d;
            e[pe[v]^1].cap += d;
        }
    }
    return ret;
}


int d[maxn];

int main()
{
    int n, m; scanf("%d %d", &n, &m);
    int u, v, c, f;
    init(n + 4);
    int ans = 0;
    for(int i = 1; i <= m; i++)
    {
        scanf("%d %d %d %d", &u, &v, &c, &f);
        d[u] -= f;
        d[v] += f;
        if(c >= f)
        {
            addedge(u, v, c - f, 1);
            addedge(u, v, INF, 2);
            addedge(v, u, f, 1);
        }
        else
        {
            ans += f - c;
            addedge(v, u, f - c, 0);
            addedge(u, v, INF, 2);
            addedge(v, u, c, 1);///c不用往下调啊啊啊
        }
    }
    addedge(n, 1, INF, 0);
    int Ss = n + 1, Tt = n + 2;
    for(int i = 1; i <= n; i++)
    {
        if(d[i] > 0) addedge(Ss, i, d[i], 0);
        if(d[i] < 0) addedge(i, Tt, -d[i], 0);
    }
    int max_flow = 0;
    ans += min_cost_flow(Ss, Tt, INF, max_flow);
    printf("%d\n", ans);
    return 0;
}
Code

 

转载于:https://www.cnblogs.com/wangwangyu/p/9856442.html

import sys from collections import deque import matplotlib.pyplot as plt import networkx as nx import numpy as np plt.rcParams['font.sans-serif'] = ['SimHei'] # 解决中文显示问题 plt.rcParams['axes.unicode_minus'] = False # 解决负号显示问题 class MinCostFlowSourceSinkVisual: def __init__(self, n, edges, source, sink, visualize=True): """ :param n: 节点数 :param edges: 边列表 [(u, v, lb, ub, cost)] :param source: 源点 :param sink: 汇点 :param visualize: 是否可视化 """ self.n = n self.source = source self.sink = sink self.original_edges = edges.copy() # 保存原始边 self.visualize = visualize self.fig, self.ax = plt.subplots(figsize=(14, 10)) self.fig.suptitle("有源汇上下界费用算法动态可视化", fontsize=16) # 初始化超级源汇 self.super_source = n self.super_sink = n + 1 self.total_nodes = n + 2 # 计算每个节点的量差 self.A = [0] * (n + 2) for u, v, lb, ub, cost in self.original_edges: self.A[u] -= lb self.A[v] += lb # 添加源汇之间无限容量的边 edges.append((sink, source, 0, float('inf'), 0)) # 创建最小费用数据结构 self.graph = [[] for _ in range(self.total_nodes)] self.dist = [float('inf')] * self.total_nodes self.vis = [False] * self.total_nodes self.pre = [-1] * self.total_nodes self.edge_info = {} # 存储边信息 self.total_cost = 0 # 总费用 # 添加图中的边 self.edge_refs = [] for i, (u, v, lb, ub, cost) in enumerate(edges): cap = ub - lb # 添加边并记录信息 self.add_edge(u, v, cap, cost, (i, lb, ub, cost, f"e{i}")) # 仅原始边(不包括后添加的sink->source边)记录在edge_refs中 if i < len(edges) - 1: # 最后一条是后添加的sink->source边 self.edge_refs.append((u, v, len(self.graph[u]) - 1, lb)) # 添加超级源汇的边 self.total_flow = 0 for i in range(n + 2): # 包含所有节点 if self.A[i] > 0: self.add_edge(self.super_source, i, self.A[i], 0, (f"S→{i}", "super_source")) self.total_flow += self.A[i] elif self.A[i] < 0: self.add_edge(i, self.super_sink, -self.A[i], 0, (f"{i}→T", "super_sink")) # 初始化可视化 if self.visualize: self.initialize_visualization() def add_edge(self, u, v, cap, cost, info=None): """添加边并存储信息""" forward = [v, cap, cost, 0, info] # [目标, 容量, 费用, 量, 信息] reverse = [u, 0, -cost, 0, None] # 反向边 forward[3] = reverse reverse[3] = forward self.graph[u].append(forward) self.graph[v].append(reverse) # 存储边信息用于可视化 if info: self.edge_info[(u, v)] = { 'capacity': cap, 'cost': cost, 'flow': 0, 'info': info } return forward def spfa(self, s, t): """SPFA算法寻找最小费用增广路径""" self.dist = [float('inf')] * self.total_nodes self.vis = [False] * self.total_nodes self.pre = [-1] * self.total_nodes self.dist[s] = 0 self.vis[s] = True queue = deque([s]) # 可视化:显示SPFA开始 if self.visualize: self.visualize_step(f"SPFA: 从超级源点S开始寻找最小费用路径") plt.pause(0.5) while queue: u = queue.popleft() self.vis[u] = False for idx, edge in enumerate(self.graph[u]): v, cap, cost, rev, info = edge if cap > 0 and self.dist[u] + cost < self.dist[v]: self.dist[v] = self.dist[u] + cost self.pre[v] = (u, idx) # 记录前驱节点和边索引 # 可视化:更新节点距离 if self.visualize: node_label = self.get_node_label(v) self.visualize_step(f"SPFA: 更新 {node_label} 距离: {self.dist[v]}") plt.pause(0.3) if not self.vis[v]: self.vis[v] = True queue.append(v) return self.dist[t] < float('inf') def min_cost_flow(self): """计算最小费用并动态可视化""" total_flow = 0 iteration = 1 while self.spfa(self.super_source, self.super_sink): # 计算增广路径上的最小容量 flow = float('inf') cur = self.super_sink path_nodes = [] while cur != self.super_source: u, idx = self.pre[cur] edge = self.graph[u][idx] path_nodes.append(cur) flow = min(flow, edge[1]) cur = u path_nodes.append(self.super_source) path_nodes.reverse() # 可视化:显示找到的增广路径 if self.visualize: path_desc = "→".join([self.get_node_label(n) for n in path_nodes]) self.visualize_step(f"找到增广路径: {path_desc}\n量: {flow}, 费用: {self.dist[self.super_sink]}") plt.pause(1.5) # 更新增广路径上的量 cur = self.super_sink path_edges = [] while cur != self.super_source: u, idx = self.pre[cur] edge = self.graph[u][idx] rev_edge = edge[3] # 更新边量 edge[1] -= flow rev_edge[1] += flow edge[4] = edge[4] or {} # 确保info存在 edge[4]['flow'] = edge[4].get('flow', 0) + flow # 更新费用 self.total_cost += flow * edge[2] # 记录路径边用于可视化 path_edges.append((u, cur)) # 更新可视化信息 if (u, cur) in self.edge_info: self.edge_info[(u, cur)]['flow'] += flow elif (cur, u) in self.edge_info: # 处理反向边 self.edge_info[(cur, u)]['flow'] -= flow cur = u # 可视化:显示量更新 if self.visualize: self.visualize_step(f"沿路径更新量: {flow}\n累计费用: {self.total_cost}") plt.pause(0.8) total_flow += flow iteration += 1 # 检查可行解 if total_flow != self.total_flow: if self.visualize: self.visualize_step(f"无可行解!\n需求量: {self.total_flow}, 实际量: {total_flow}") plt.pause(3.0) return None, None # 计算原图中每条边的实际量 flows = [] for u, v, idx, lb in self.edge_refs: # 跳过最后添加的sink->source边 if u == self.sink and v == self.source: continue edge = self.graph[u][idx] actual_flow = lb + edge[1] # 实际量 = 下界 + 残余网络中的剩余容量 flows.append(actual_flow) if self.visualize: self.visualize_final_flow(flows) plt.pause(5.0) return flows, self.total_cost def get_node_label(self, node): """获取节点标签""" if node == self.super_source: return "S" elif node == self.super_sink: return "T" elif node == self.source: return f"源点({node})" elif node == self.sink: return f"汇点({node})" else: return f"{node}" def get_edge_description(self, u, v): """获取边的描述信息""" if u == self.super_source: return f"S → {v}" elif v == self.super_sink: return f"{u} → T" elif u == self.source and v == self.sink: return f"{u}→{v} (源汇边)" elif (u, v) in self.edge_info: info = self.edge_info[(u, v)]['info'] if isinstance(info, tuple) and len(info) > 3: return f"{u} → {v} ({info[4]})" return f"{u} → {v}" def initialize_visualization(self): """初始化可视化布局""" self.G = nx.DiGraph() # 添加节点 for i in range(self.n): self.G.add_node(i, label=f"{i}") self.G.add_node(self.super_source, label="S") self.G.add_node(self.super_sink, label="T") # 添加边 for u in range(self.total_nodes): for edge in self.graph[u]: v, cap, cost, _, info = edge if cap > 0: # 只添加正向边 self.G.add_edge(u, v, capacity=cap, cost=cost, flow=0) # 创建环形布局 self.pos = {} # 普通节点布置在圆上 angles = np.linspace(0, 2 * np.pi, self.n, endpoint=False) for i in range(self.n): angle = angles[i] self.pos[i] = (np.cos(angle), np.sin(angle)) # 特殊节点位置 self.pos[self.source] = (0, 1.2) # 源点在上方 self.pos[self.sink] = (0, -1.2) # 汇点在下方 self.pos[self.super_source] = (-1.5, 0) # 超级源点在左侧 self.pos[self.super_sink] = (1.5, 0) # 超级汇点在右侧 # 初始绘图 self.ax.clear() # 节点颜色:普通节点-浅蓝,源汇点-浅绿,超级源汇-浅红 node_colors = [] for node in self.G.nodes(): if node == self.source or node == self.sink: node_colors.append('lightgreen') elif node == self.super_source or node == self.super_sink: node_colors.append('salmon') else: node_colors.append('lightblue') nx.draw_networkx_nodes(self.G, self.pos, node_size=800, node_color=node_colors) nx.draw_networkx_labels(self.G, self.pos, labels={n: d['label'] for n, d in self.G.nodes(data=True)}) # 绘制边 self.edge_collection = nx.draw_networkx_edges( self.G, self.pos, arrowstyle='->', arrowsize=20, edge_color='gray', width=1, ax=self.ax ) # 初始化边标签 self.edge_labels = {} for u, v in self.G.edges(): self.edge_labels[(u, v)] = self.ax.text(0, 0, "", fontsize=8, ha='center', va='center') self.ax.set_title("初始化网络", fontsize=14) self.ax.set_axis_off() plt.tight_layout() plt.pause(2.0) def visualize_step(self, message): """可视化当前步骤""" self.ax.clear() # 节点颜色 node_colors = [] for node in self.G.nodes(): if node == self.source or node == self.sink: node_colors.append('lightgreen') elif node == self.super_source or node == self.super_sink: node_colors.append('salmon') else: node_colors.append('lightblue') # 绘制节点 nx.draw_networkx_nodes(self.G, self.pos, node_size=800, node_color=node_colors) nx.draw_networkx_labels(self.G, self.pos, labels={n: d['label'] for n, d in self.G.nodes(data=True)}) # 绘制边并设置颜色和宽度 edge_colors = [] edge_widths = [] for u, v in self.G.edges(): # 获取当前边的状态 cap = self.G[u][v]['capacity'] flow = self.edge_info.get((u, v), {}).get('flow', 0) # 计算饱和度 saturation = flow / cap if cap > 0 else 0 # 使用颜色表示饱和度 edge_colors.append(plt.cm.RdYlGn(saturation)) # 使用宽度表示量 edge_widths.append(1 + 3 * saturation) # 绘制边 nx.draw_networkx_edges( self.G, self.pos, arrowstyle='->', arrowsize=20, edge_color=edge_colors, width=edge_widths, ax=self.ax ) # 更新边标签 for (u, v), text in self.edge_labels.items(): # 获取边信息 cap = self.G[u][v]['capacity'] cost = self.G[u][v]['cost'] flow = self.edge_info.get((u, v), {}).get('flow', 0) # 特殊边处理 if u == self.super_source or v == self.super_sink: label = f"{flow}/{cap}\n费用:0" else: # 获取原始边信息 info = self.edge_info.get((u, v), {}).get('info', None) if info and isinstance(info, tuple): _, lb, ub, cost_val, name = info actual_flow = lb + flow label = f"{name}: {actual_flow}/{ub}\n费用:{cost_val}\n[{lb},{ub}]" else: label = f"{flow}/{cap}\n费用:{cost}" # 计算边的中点位置 x = (self.pos[u][0] + self.pos[v][0]) / 2 y = (self.pos[u][1] + self.pos[v][1]) / 2 # 更新文本位置和内容 text.set_position((x, y)) text.set_text(label) self.ax.add_artist(text) # 显示当前信息 self.ax.set_title(f"{message}\n总费用: {self.total_cost}", fontsize=14) self.ax.set_axis_off() plt.tight_layout() plt.draw() def visualize_final_flow(self, flows): """可视化最终可行分配(仅显示原图边)""" self.ax.clear() # 创建仅包含原图节点和边的子图 H = nx.DiGraph() for i in range(self.n): H.add_node(i, label=f"{i}") # 添加原图边(排除最后添加的sink->source边) for i, (u, v, lb, ub, cost) in enumerate(self.original_edges): if i >= len(flows): continue H.add_edge(u, v, flow=flows[i], lb=lb, ub=ub, cost=cost, name=f"e{i}") # 使用原布局,但只保留原图节点的位置 pos = {k: v for k, v in self.pos.items() if k in H.nodes()} # 绘制节点 node_colors = ['lightgreen' if node == self.source or node == self.sink else 'lightblue' for node in H.nodes()] nx.draw_networkx_nodes(H, pos, node_size=800, node_color=node_colors) nx.draw_networkx_labels(H, pos) # 绘制边并设置颜色和宽度 edge_colors = [] edge_widths = [] for u, v in H.edges(): flow = H[u][v]['flow'] ub = H[u][v]['ub'] saturation = flow / ub edge_colors.append(plt.cm.RdYlGn(saturation)) edge_widths.append(1 + 3 * saturation) nx.draw_networkx_edges( H, pos, arrowstyle='->', arrowsize=20, edge_color=edge_colors, width=edge_widths, ax=self.ax ) # 添加边标签 edge_labels = {} for u, v in H.edges(): flow = H[u][v]['flow'] lb = H[u][v]['lb'] ub = H[u][v]['ub'] cost = H[u][v]['cost'] name = H[u][v]['name'] edge_labels[(u, v)] = f"{name}: {flow}\n费用:{cost}\n[{lb},{ub}]" nx.draw_networkx_edge_labels(H, pos, edge_labels=edge_labels, font_size=8) self.ax.set_title(f"最小费用分配结果(总费用: {self.total_cost})", fontsize=14) self.ax.set_axis_off() plt.tight_layout() plt.draw() def min_cost_flow_visual(n, edges, source, sink): """有源汇上下界费用求解可视化""" # 创建可视化实例 mcf_visual = MinCostFlowSourceSinkVisual(n, edges, source, sink, visualize=True) # 计算最小费用 flows, total_cost = mcf_visual.min_cost_flow() if flows is None: print("无可行解") return None, None print("\n各边实际量分配和费用:") for i, (u, v, lb, ub, cost) in enumerate(edges[:-1]): # 排除最后添加的sink->source边 print(f"边 {u}→{v} ({lb},{ub}): 量={flows[i]}, 费用={cost}") # 计算源点到汇点的总量 source_flow = sum(flows[i] for i, (u, v, _, _, _) in enumerate(edges) if u == source) sink_flow = sum(flows[i] for i, (u, v, _, _, _) in enumerate(edges) if v == sink) print(f"\n源点({source})总输出量: {source_flow}") print(f"汇点({sink})总输入量: {sink_flow}") print(f"总费用: {total_cost}") plt.show() # 保持窗口打开 return flows, total_cost if __name__ == "__main__": # 15节点有可行解的网络示例 - 简化版 print("=" * 50) print("15节点网络的有源汇上下界费用计算 (保证有可行解)") # 简化设计:确保网络平衡 n = 15 edges = [ # 源点→核心节点 (u, v, lb, ub, cost) (0, 1, 5, 10, 2), (0, 2, 5, 10, 3), # 核心环状结构 (1, 2, 0, 5, 1), (2, 3, 2, 8, 4), (3, 4, 2, 8, 2), (4, 1, 0, 5, 1), # 核心→中间节点 (1, 5, 1, 4, 3), (2, 6, 1, 4, 2), (3, 7, 1, 4, 4), (4, 8, 1, 4, 3), # 中间层平衡结构 (5, 6, 0, 5, 1), (6, 7, 0, 5, 2), (7, 8, 0, 5, 1), (8, 5, 0, 5, 3), # 中间→汇点 (5, 14, 3, 6, 5), (6, 14, 3, 6, 4), (7, 14, 2, 5, 3), (8, 14, 2, 5, 2), # 连接外围节点 (1, 9, 0, 3, 2), (2, 10, 0, 3, 3), (3, 11, 0, 3, 1), (4, 12, 0, 3, 2), (9, 13, 0, 3, 4), (10, 13, 0, 3, 2), (11, 13, 0, 3, 3), (12, 13, 0, 3, 1), (13, 14, 0, 5, 2) # 汇点入口 ] # 设置源点和汇点 source = 0 # 节点0作为源点 sink = 14 # 节点14作为汇点 # 计算并可视化最小费用 flows, total_cost = min_cost_flow_visual(n, edges, source, sink) C:\Users\25827\.conda\envs\torch\python.exe C:\Users\25827\Desktop\图论代码\有源汇上下界费用.py ================================================== 15节点网络的有源汇上下界费用计算 (保证有可行解) Traceback (most recent call last): File "C:\Users\25827\Desktop\图论代码\有源汇上下界费用.py", line 516, in <module> flows, total_cost = min_cost_flow_visual(n, edges, source, sink) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\Users\25827\Desktop\图论代码\有源汇上下界费用.py", line 442, in min_cost_flow_visual flows, total_cost = mcf_visual.min_cost_flow() ^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\Users\25827\Desktop\图论代码\有源汇上下界费用.py", line 165, in min_cost_flow edge[4]['flow'] = edge[4].get('flow', 0) + flow ^^^^^^^^^^^ AttributeError: 'tuple' object has no attribute 'get' 进程已结束,退出代码为 1 给出修改后的完整代码
最新发布
06-15
评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值