def main():
# test
num_nodes = 8
G = r_tree(num_nodes)
plot_graph(G, 0, 'environment_debug_graph.png')
env = n_environment(G, [0], 1)
for x in range(1, num_nodes):
print(env.step(x))
env.reset()
print('Reset env =========================')
while not env.done:
print(env.step(random.randint(0, num_nodes - 1)))
print()
def sanity_test():
# Run a quick sanity check
num_nodes = 7
G = r_tree(num_nodes)
plot_graph(G, get_root(G), 'environment_debug_graph.png')
env = environment(G, [get_root(G)], 1)
while not env.done:
print(env.step(random.randint(0, 10)))
print()
env.reset()
print('Reset env =========================')
while not env.done:
print(env.step(random.randint(0, 10)))
print()
def main():
num_nodes = 7
resources = 1
G = r_tree(num_nodes)
plot_graph(G, 0, 'environment_debug_graph.png')
utils = [0 for _ in range(num_nodes)]
for key, val in nx.get_node_attributes(G, 'util').items():
utils[key] = val
demand = [0 for _ in range(num_nodes)]
for key, val in nx.get_node_attributes(G, 'demand').items():
demand[key] = val
N = len(utils)
S = int(np.ceil(np.sum(demand) / resources))
obj_func = create_u_t(G, utils, demand, resources)
print(obj_func(np.zeros(N * S)))
def generate_graph(nodes=20,
utils=[1, 4],
demands=[1, 2],
load_dir=None,
type='random_tree',
seed=None):
# Generate random tree
if type == 'random_tree':
if load_dir:
graph = nx.read_gpickle(
'experiments/{0}_rtree.gpickle'.format(nodes))
else:
graph = r_tree(nodes, utils, demands)
save = 'experiments/{0}_rtree.txt'.format(nodes)
real_node_num = nodes
# Generate random graph by adding edges according to some p-value (0.2)
elif type == 'random_graph':
if load_dir:
graph = nx.read_gpickle(
'experiments/{0}_rgraph.gpickle'.format(nodes))
else:
graph = r_graph(nodes, 0.2, utils, demands, seed)
save = 'experiments/{0}_rgraph.txt'.format(nodes)
real_node_num = nodes
# Generate random nodes x nodes 2-d grid graph
elif type == 'grid':
if load_dir:
graph = nx.read_gpickle(
'experiments/{0}x{0}.gpickle'.format(nodes))
else:
graph = r_2d_graph(nodes, nodes, utils, demands)
save = 'experiments/{0}x{0}.txt'.format(nodes)
real_node_num = nodes**2
# Generate adversarial examples for the ratio heuristic
elif type == 'adversarial':
graph = adv_graph(nodes, utils, demands)
save = 'experiments/{0}_adv_graph.txt'.format(nodes)
real_node_num = nodes
# Read a normal gml file and randomly pick utils, demands
elif type == 'gml':
# WARNING: Turn off fix_nodes_around_adv for normal use. This fixes the
# nodes adjacent to vertex(num_nodes - 2) to be bad for the ratio heuristic.
# For use in conjunction with 'gml_adversarial'.
graph = read_gml(load_dir, utils, demands, fix_nodes_around_adv=False)
save = 'experiments/{0}.txt'.format('gml')
real_node_num = len(graph)
# Read a gml file and randomly pick utils, demands but also
# embed an adversarial example somewhere in the graph.
elif type == 'gml_adversarial':
graph = read_gml_adversarial(load_dir, utils, demands)
save = 'experiments/{0}_adv_graph.txt'.format('gml')
real_node_num = len(graph)
elif type == 'gnp_adversarial':
graph = gnp_adversarial(nodes, utils, demands)
save = 'experiments/{0}_gnp_adv_graph.txt'.format(nodes)
real_node_num = nodes
else:
raise NotImplementedError
# Real number of nodes may be different from input node num (in the case of grid graph)
return graph, save, real_node_num
def main():
tree = r_tree(nodes=8)
root = get_root(tree)
ratio_heuristic(tree, [root], 1)