-
Notifications
You must be signed in to change notification settings - Fork 0
/
Graph.py
199 lines (161 loc) · 7.53 KB
/
Graph.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
from __future__ import annotations
from enum import Enum
from itertools import product
import math
from typing import Dict, Set, Tuple, Optional
from tqdm import tqdm
import random
import logging
import pickle
import sys
#Set this to False to disable debug prints
DEBUG = False
INFO = True
# Set up logging
if DEBUG:
logging.basicConfig(stream=sys.stdout, level=logging.DEBUG)
if INFO:
logging.basicConfig(stream=sys.stdout, level=logging.INFO)
class Player(Enum):
MIN = 1
MAX = 2
class GenerationStrategy(Enum):
BELLMAN_FORD = "bellman_ford"
INCREMENTAL_BELLMAN_FORD = "incremental_bellman_ford"
NONE = 'none'
class Arena:
def __init__(self,
num_nodes: int = 10,
edge_probability: float = 0.01,
seed: int | None = None):
self.nodes = set(range(num_nodes))
self.random = random.Random(seed)
self.value_mapping: Dict[int, float] = {i: 0 for i in range(num_nodes)}
self.edges_mapping: Dict[int, Set[Tuple[int, int, float]]] = {i: set() for i in range(num_nodes)}
self.edges = set()
self.num_nodes = num_nodes
self.edge_probability = edge_probability
self.max_weight = 10
self.weight_range = (-self.max_weight, self.max_weight)
self.distances: Dict[int, Optional[float]] = {node: 0 for node in self.nodes}
self.considered: Set[Tuple[int, Tuple[int, int, float]]]= set()
self.considered_edges_bf: Set[Tuple[int, int, float]] = set()
self.fast_edges: Dict[int, Dict[int, float]] = {i: {} for i in range(num_nodes)}
self.player_mapping: Dict[int, Player] = self._assign_players(equal=True)
self.ingoing_edges: Dict[int, Set[Tuple[int, int, float]]] = {i: set() for i in range(num_nodes)}
def __repr__(self):
return f"Nodes: {self.nodes}\nEdges: {self.edges}"
def __str__(self):
return f"Nodes: {self.nodes}\nEdges: {self.edges}"
def save(self, save_path: str = "arena.pkl") -> None:
with open(save_path, 'wb') as f:
pickle.dump(self, f)
def load(self, pickle_file: str) -> Arena:
with open(pickle_file, 'rb') as f:
arena = pickle.load(f)
return arena
def _assign_players(self, equal: bool = False) -> Dict[int, Player]:
"""
Assign players to the nodes.
"""
if equal:
players = [Player.MAX] * (self.num_nodes // 2) + [Player.MIN] * (self.num_nodes // 2)
self.random.shuffle(players)
player_mapping = {i: player for i, player in enumerate(players)}
else:
player_mapping = {i: self.random.choice([Player.MIN, Player.MAX]) for i in range(self.num_nodes)}
return player_mapping
def safely_update(self,
node: int,
edge: Tuple[int, int, float],
strategy: GenerationStrategy):
"""
Update the arena with the new edge and backtrack if a negative cycle is detected.
"""
if strategy == GenerationStrategy.NONE.value:
negative_cycles = False
elif strategy == GenerationStrategy.BELLMAN_FORD.value:
negative_cycles = self.bellman_ford(nodes=[*self.nodes, node], edges=[*self.edges, edge])
elif strategy == GenerationStrategy.INCREMENTAL_BELLMAN_FORD.value:
negative_cycles = self.bellman_ford_incremental(edge)
else:
raise ValueError(f"Invalid strategy: {strategy}")
if negative_cycles:
return
# assert node == edge[0]
self.edges_mapping[node].add(edge)
self.ingoing_edges[edge[1]].add(edge)
def generate(self,
strategy: GenerationStrategy = GenerationStrategy.INCREMENTAL_BELLMAN_FORD):
pbar = tqdm(total=self.num_nodes ** 2, desc=f"Creating graph (n = {self.num_nodes}, p = {self.edge_probability})")
update_delta = round(math.sqrt(pbar.total))
for i, (origin, dest) in enumerate(product(self.nodes, repeat=2)):
if i % update_delta == 0:
pbar.update(update_delta)
if self.random.random() < self.edge_probability:
weight = self.random.uniform(*self.weight_range)
# Avoid self loops with negative weight
if not (origin == dest and weight < 0):
edge = (origin, dest, weight)
self.safely_update(node=origin,
edge=edge,
strategy=strategy)
pbar.close()
def bellman_ford_incremental(self, new_edge: Tuple[int, int, float]) -> bool:
"""
A very efficient implementation of the Bellman-Ford algorithm that only checks for negative cycles related to the new edge.
It uses the fast_edges dictionary to keep track of the edges and their weights, in order to avoid iterating over all the edges.
"""
#TODO: this creates a valid graph, but sometimes it has some false positives
# Add the new edge
self.edges.add(new_edge)
self.fast_edges[new_edge[0]][new_edge[1]] = new_edge[2]
new_distance_0 = self.distances.get(new_edge[0], 0) + new_edge[2]
new_distance_1 = self.distances.get(new_edge[1], float('inf'))
# previous_distance_0 = self.distances.get(new_edge[0], None)
previous_distance_1 = self.distances.get(new_edge[1], None)
# Relax edges related to the new edge
if new_distance_0 < new_distance_1:
new_distance_1 = new_distance_0
self.distances[new_edge[1]] = new_distance_1
origin_to = self.fast_edges[new_edge[0]] #this is a dict that maps the destination to the weight
dest_to = self.fast_edges[new_edge[1]] #this is a dict that maps the origin to the weight
edges = {(new_edge[0], dest, weight) for dest, weight in origin_to.items()} | {(new_edge[1], origin, weight) for origin, weight in dest_to.items()}
for edge in edges:
if self.distances[edge[0]] + edge[2] < self.distances.get(edge[1], float('inf')):
self.edges.remove(new_edge)
self.fast_edges[new_edge[0]].pop(new_edge[1])
self.distances[new_edge[1]] = previous_distance_1
return True
return False # No negative cycle found
def bellman_ford(self,
nodes: Optional[Set[int]] = None,
edges: Optional[Set[Tuple[int, int, float]]] = None) -> bool:
"""
Detect negative cycles using Bellman-Ford algorithm.
"""
if nodes is None:
nodes = self.nodes
if edges is None:
edges = self.edges
distances: Dict[int, float] = {node: 0 for node in nodes}
# Relax edges repeatedly
for _ in range(len(nodes) - 1):
for edge in edges:
if distances[edge[0]] + edge[2] < distances.get(edge[1], float('inf')):
distances[edge[1]] = distances[edge[0]] + edge[2]
# Check for negative cycles
for edge in edges:
if distances[edge[0]] + edge[2] < distances.get(edge[1], float('inf')):
return True # Negative cycle found
return False # No negative cycle found
def get_outgoing_edges(self, node: int) -> Set[Tuple[int, int, float]]:
"""
Get the outgoing edges of a node.
"""
return self.edges_mapping[node]
def get_node_degree(self, node: int) -> int:
"""
Get the degree of a node.
"""
return len(self.get_outgoing_edges(node))