deastern/lead
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

更新readme，将mc配置文件修改为多目标优化

Browse Source
This commit is contained in:
yangyudong 2025-04-16 01:32:53 +08:00
parent 6d909b5e4f
commit f576f8d02e
5 changed files with 185 additions and 12 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

86
README.md
View File

@ -1,3 +1,85 @@
# lead
# LEAD (LLM-Evolved Algorithm Development)
llm驱动的算法开发流程
LLM驱动的算法进化开发框架，支持单目标和多目标优化问题的算法自动进化。
## 单目标算法进化
### 1. 配置要求
在问题目录下的`problem_config.json`中配置：
```json
{
"multi_objective": false,
"description": "问题描述",
"function_name": "函数名",
"input_format": "输入格式",
"output_format": "输出格式",
"evolution_params": {
"algorithm": "TU", # 或"DE"
"population_size": 8,
"generations": 10
}
}
```
### 2. 运行流程
1. 初始化种群：LLM根据问题描述生成初始算法代码
2. 评估：计算每个个体的适应度(单目标值)
3. 进化循环：
- 选择：锦标赛选择或差分进化选择
- 交叉：代码片段重组
- 变异：LLM引导的代码变异
4. 输出：最优算法代码和适应度曲线
### 3. 运行命令
```bash
lead-run --problem [问题名] [--generations 代数] [--population 种群大小]
```
可选参数：
- `--generations`: 覆盖默认进化代数
- `--population`: 覆盖默认种群大小
## 多目标算法进化
### 1. 配置要求
```json
{
"multi_objective": true,
"objective_names": ["目标1", "目标2"],
"description": "问题描述",
"evolution_params": {
"algorithm": "MO",
"population_size": 10,
"generations": 20
}
}
```
### 2. 运行流程
1. 初始化种群：LLM生成初始多目标算法
2. 评估：计算每个个体的多目标适应度向量
3. 进化循环：
- 非支配排序
- 拥挤度计算
- 精英保留
4. 输出：Pareto前沿算法集合
### 3. 运行命令
```bash
lead-run --problem [问题名] [--generations 代数] [--population 种群大小]
```
可选参数同上
## 配置说明
- `problem_config.json`必须包含：
- 问题描述
- 输入输出格式
- 进化参数
- 可选配置：
- LLM参数
- 评估超时时间
## 结果输出
- 单目标：`evolution_history/[时间戳]/`目录保存各代最优解
- 多目标：`multi_objective_history/[时间戳]/`保存Pareto前沿

23
src/lead/problems/mc/evaluation.py
View File

@ -2,16 +2,18 @@ import numpy as np
import time
from typing import Callable, Dict
def evaluate(coloring_func: Callable[[np.ndarray], np.ndarray], adj_matrix: np.ndarray) -> float:
"""评估图着色算法的性能
def evaluate(coloring_func: Callable[[np.ndarray], np.ndarray], adj_matrix: np.ndarray) -> Dict[str, float]:
"""评估图着色算法的多目标性能
Args:
coloring_func: 图着色函数
adj_matrix: 图的邻接矩阵
Returns:
float: 适应度分数 = 运行时间 * 使用的颜色数
如果解不合法(存在相邻节点同色)则返回无穷大
Dict[str, float]: 包含两个目标值的字典:
- color_count: 使用的颜色数量
- execution_time: 运行时间(秒)
如果解不合法(存在相邻节点同色)则返回无穷大
"""
try:
# 计时并执行图着色
@ -26,17 +28,20 @@ def evaluate(coloring_func: Callable[[np.ndarray], np.ndarray], adj_matrix: np.n
for j in range(n):
if adj_matrix[i][j] == 1 and colors[i] == colors[j]:
print(f"非法解:节点{i}和{j}相邻但颜色相同")
return float('inf')
return {
"color_count": float('inf'),
"execution_time": float('inf')
}
# 计算使用的颜色数
num_colors = len(set(colors))
# 计算适应度分数
fitness = num_colors + execution_time
print(f"执行时间:{execution_time:.4f}秒, 使用颜色数:{num_colors}")
# print(f"适应度分数:{fitness:.4f}")
return num_colors
return {
"color_count": num_colors,
"execution_time": execution_time
}
except Exception as e:
print(f"评估过程出错: {str(e)}")

1
src/lead/problems/mc/multi_objective_evolution.log Normal file
View File

@ -0,0 +1 @@
Generation 0000 | Pareto Front Size: 1 | Timestamp: 2025-04-16T01:20:59.478238

83
src/lead/problems/mc/multi_objective_history/20250416_010845/generation_0.json Normal file
View File

@ -0,0 +1,83 @@
{
"generation": 0,
"population": [
{
"code": "import numpy as np\n\ndef graph_coloring(adj_matrix):\n n = adj_matrix.shape[0]\n colors = [-1] * n\n available = [False] * n\n \n def get_neighbor_colors(node):\n for neighbor in range(n):\n if adj_matrix[node][neighbor] == 1 and colors[neighbor] != -1:\n available[colors[neighbor]] = True\n\n def assign_color(node):\n get_neighbor_colors(node)\n color = 0\n while color < n:\n if not available[color]:\n colors[node] = color\n break\n color += 1\n for neighbor in range(n):\n if adj_matrix[node][neighbor] == 1 and colors[neighbor] != -1:\n available[colors[neighbor]] = False\n \n for node in range(n):\n if colors[node] == -1:\n assign_color(node)\n available = [False] * n\n\n return colors",
"generation": 0,
"fitnesses": {
"color_count": 31,
"execution_time": 1.0018470287322998
}
},
{
"code": "import numpy as np\nimport random\n\ndef graph_coloring(adj_matrix):\n n = adj_matrix.shape[0]\n colors = [-1] * n\n \n vertex_order = list(range(n))\n random.shuffle(vertex_order)\n \n for vertex in vertex_order:\n available_colors = [True] * n\n for neighbor in range(n):\n if adj_matrix[vertex][neighbor] == 1 and colors[neighbor] != -1:\n available_colors[colors[neighbor]] = False\n \n color = 0\n while color < n and not available_colors[color]:\n color += 1\n \n colors[vertex] = color\n \n return colors",
"generation": 0,
"fitnesses": {
"color_count": 32,
"execution_time": 1.5306849479675293
}
},
{
"code": "import numpy as np\nimport random\n\ndef graph_coloring(adj_matrix):\n n = len(adj_matrix)\n colors = [-1] * n\n available_colors = [True] * n\n\n def greedy_coloring():\n colors[0] = 0\n\n for u in range(1, n):\n available_colors[:] = [True] * n\n \n for v in range(n):\n if adj_matrix[u][v] == 1 and colors[v] != -1:\n available_colors[colors[v]] = False\n \n for color in range(n):\n if available_colors[color]:\n colors[u] = color\n break\n\n greedy_coloring()\n \n def optimize_colors():\n color_count = max(colors) + 1\n color_mapping = {i: [] for i in range(color_count)}\n \n for vertex in range(n):\n if colors[vertex] != -1:\n color_mapping[colors[vertex]].append(vertex)\n \n for color in range(color_count):\n if not color_mapping[color]:\n continue\n for new_color in range(color + 1, color_count):\n if new_color not in color_mapping:\n continue\n is_codable = True\n for v1 in color_mapping[color]:\n for v2 in color_mapping[new_color]:\n if adj_matrix[v1][v2] == 1:\n is_codable = False\n break\n if not is_codable:\n break\n if is_codable:\n for v in color_mapping[new_color]:\n colors[v] = color\n del color_mapping[new_color]\n break\n\n optimize_colors()\n \n return colors",
"generation": 0,
"fitnesses": {
"color_count": 31,
"execution_time": 1.539431095123291
}
},
{
"code": "import numpy as np\nimport random\n\ndef graph_coloring(adj_matrix):\n n = len(adj_matrix)\n colors = [-1] * n\n available_colors = [False] * n\n \n def find_available_color(vertex):\n for i in range(n):\n available_colors[i] = True\n \n for neighbor in range(n):\n if adj_matrix[vertex][neighbor] == 1 and colors[neighbor] != -1:\n available_colors[colors[neighbor]] = False\n \n for color in range(n):\n if available_colors[color]:\n return color\n return n\n\n for vertex in range(n):\n colors[vertex] = find_available_color(vertex)\n \n def optimize_coloring():\n unique_colors = sorted(set(colors))\n color_map = {old_color: new_color for new_color, old_color in enumerate(unique_colors)}\n for i in range(n):\n colors[i] = color_map[colors[i]]\n\n optimize_coloring()\n \n return colors",
"generation": 0,
"fitnesses": {
"color_count": 31,
"execution_time": 2.813322067260742
}
},
{
"code": "import numpy as np\nimport random\n\ndef graph_coloring(adj_matrix):\n n = adj_matrix.shape[0]\n colors = [-1] * n\n \n def get_available_colors(vertex):\n available_colors = [True] * n\n for neighbor in range(n):\n if adj_matrix[vertex][neighbor] == 1 and colors[neighbor] != -1:\n available_colors[colors[neighbor]] = False\n return available_colors\n\n vertex_order = list(range(n))\n random.shuffle(vertex_order)\n \n for vertex in vertex_order:\n available_colors = get_available_colors(vertex)\n \n color = 0\n while color < n and not available_colors[color]:\n color += 1\n \n if color == n:\n min_color_used = min([c for c in colors if c != -1], default=0)\n color = min_color_used + 1\n \n colors[vertex] = color\n \n return colors",
"generation": 0,
"fitnesses": {
"color_count": 31,
"execution_time": 2.3673529624938965
}
},
{
"code": "import numpy as np\nimport random\n\ndef graph_coloring(adj_matrix):\n n = adj_matrix.shape[0]\n colors = [-1] * n\n available_colors = [True] * n\n \n def assign_colors():\n for node in range(n):\n if colors[node] == -1:\n reset_available()\n for neighbor in range(n):\n if adj_matrix[node][neighbor] == 1 and colors[neighbor] != -1:\n available_colors[colors[neighbor]] = False\n colors[node] = get_first_available_color()\n \n def reset_available():\n for i in range(n):\n available_colors[i] = True\n\n def get_first_available_color():\n for color in range(n):\n if available_colors[color]:\n return color\n return n\n\n def local_search():\n for node in range(n):\n current_color = colors[node]\n for try_color in range(n):\n if try_color != current_color and is_valid_color_assignment(node, try_color):\n colors[node] = try_color\n break\n\n def is_valid_color_assignment(node, color):\n for neighbor in range(n):\n if adj_matrix[node][neighbor] == 1 and colors[neighbor] == color:\n return False\n return True\n\n assign_colors()\n \n for _ in range(n):\n local_search()\n \n return colors",
"generation": 0,
"fitnesses": {
"color_count": Infinity,
"execution_time": Infinity
}
},
{
"code": "import numpy as np\nimport random\n\ndef graph_coloring(adj_matrix):\n n = len(adj_matrix)\n colors = [-1] * n\n available_colors = [True] * n\n\n def is_color_possible(vertex, color):\n for neighbor in range(n):\n if adj_matrix[vertex][neighbor] == 1 and colors[neighbor] == color:\n return False\n return True\n\n def greedy_coloring():\n for vertex in range(n):\n available_colors[:] = [True] * n\n for neighbor in range(n):\n if adj_matrix[vertex][neighbor] == 1 and colors[neighbor] != -1:\n available_colors[colors[neighbor]] = False\n for color in range(n):\n if available_colors[color]:\n colors[vertex] = color\n break\n \n greedy_coloring()\n\n def local_search():\n improved = True\n while improved:\n improved = False\n for vertex in range(n):\n current_color = colors[vertex]\n conflict_colors = set()\n for neighbor in range(n):\n if adj_matrix[vertex][neighbor] == 1:\n conflict_colors.add(colors[neighbor])\n for new_color in range(len(conflict_colors) + 1):\n if new_color not in conflict_colors:\n if new_color != current_color:\n colors[vertex] = new_color\n improved = True\n break\n\n local_search()\n\n return colors",
"generation": 0,
"fitnesses": {
"color_count": Infinity,
"execution_time": Infinity
}
},
{
"code": "import numpy as np\nimport random\n\ndef graph_coloring(adj_matrix):\n n = len(adj_matrix)\n colors = [-1] * n\n available_colors = [True] * n\n\n def greedy_coloring():\n colors[0] = 0\n\n for u in range(1, n):\n available_colors[:] = [True] * n\n \n for v in range(n):\n if adj_matrix[u][v] == 1 and colors[v] != -1:\n available_colors[colors[v]] = False\n \n for color in range(n):\n if available_colors[color]:\n colors[u] = color\n break\n\n greedy_coloring()\n \n def optimize_colors():\n color_count = max(colors) + 1\n color_mapping = {i: [] for i in range(color_count)}\n \n for vertex in range(n):\n if colors[vertex] != -1:\n color_mapping[colors[vertex]].append(vertex)\n \n merged = True\n while merged:\n merged = False\n for color in range(color_count):\n if not color_mapping[color]:\n continue\n for new_color in range(color + 1, color_count):\n if new_color not in color_mapping:\n continue\n is_codable = True\n for v1 in color_mapping[color]:\n for v2 in color_mapping[new_color]:\n if adj_matrix[v1][v2] == 1:\n is_codable = False\n break\n if not is_codable:\n break\n if is_codable:\n for v in color_mapping[new_color]:\n colors[v] = color\n del color_mapping[new_color]\n merged = True\n color_count -= 1\n break\n if merged:\n break\n\n optimize_colors()\n \n return colors",
"generation": 0,
"fitnesses": {
"color_count": 31,
"execution_time": 2.5284109115600586
}
}
],
"pareto_front": [
{
"code": "import numpy as np\n\ndef graph_coloring(adj_matrix):\n n = adj_matrix.shape[0]\n colors = [-1] * n\n available = [False] * n\n \n def get_neighbor_colors(node):\n for neighbor in range(n):\n if adj_matrix[node][neighbor] == 1 and colors[neighbor] != -1:\n available[colors[neighbor]] = True\n\n def assign_color(node):\n get_neighbor_colors(node)\n color = 0\n while color < n:\n if not available[color]:\n colors[node] = color\n break\n color += 1\n for neighbor in range(n):\n if adj_matrix[node][neighbor] == 1 and colors[neighbor] != -1:\n available[colors[neighbor]] = False\n \n for node in range(n):\n if colors[node] == -1:\n assign_color(node)\n available = [False] * n\n\n return colors",
"generation": 0,
"fitnesses": {
"color_count": 31,
"execution_time": 1.0018470287322998
}
}
],
"objective_names": [
"color_count",
"execution_time"
]
}

4
src/lead/problems/mc/problem_config.json
View File

@ -3,6 +3,8 @@
"function_name": "graph_coloring",
"input_format": "一个numpy数组adj_matrix，表示图的邻接矩阵（n×n的0-1整数矩阵，1表示两点相连，0表示不相连）",
"output_format": "返回一个长度为n的整数数组colors，表示每个顶点的颜色编号（从0开始的整数）",
"multi_objective": true,
"objective_names": ["color_count", "execution_time"],
"evaluation_timeout": 120,
"llm_config": {
"api_key": "sk-5wHuJRuC2KDMINeYERru0zCCMrRTzmkrOMwpvBQr4EULV0Tv",
@ -12,7 +14,7 @@
"max_tokens": 8192
},
"evolution_params": {
"algorithm": "TU",
"algorithm": "MO",
"population_size": 8,
"generations": 10,
"mutation_rate": 0.5,