def main():
for x in range(12, 15):
agents, tasks = generate_painter(x, x, 3)
for p in [0, 0.05, 0.1, 0.15, 0.2]:
results = []
for n in range(10):
# We give back m tasks in a giveback phase, and each iteration we assign one task,
# so for m agents we should have a giveback probability < 1/m
auctioneer = Local_Auctioneer(deepcopy(agents), deepcopy(tasks), p)#1 / ( * len(agents)))
local_schedule = auctioneer.optimize_schedule()
local_pr = auctioneer.assignment_probability(local_schedule)
global_auctioneer = Auctioneer(deepcopy(agents), deepcopy(tasks), p)
global_schedule = global_auctioneer.optimize_schedule()
global_pr = auctioneer.assignment_probability(global_schedule)
#print(",".join([str(x) for x in (x, p, local_pr, global_pr)]))
results += [(local_pr, global_pr)]
local_results, global_results = zip(*results)
local_mean = np.mean(local_results)
global_mean = np.mean(global_results)
# https://stackoverflow.com/questions/15033511/compute-a-confidence-interval-from-sample-data
local_min_ci, local_max_ci = st.t.interval(0.95, len(local_results) - 1, loc=np.mean(local_results), scale=st.sem(local_results))
global_min_ci, global_max_ci = st.t.interval(0.95, len(global_results) - 1, loc=np.mean(global_results), scale=st.sem(global_results))
A = [x, p, local_mean, global_mean, local_min_ci, local_max_ci, global_min_ci, global_max_ci]
print(",".join([str(z) for z in A]))
def main():
agents, tasks = generate_painter(X, Y, 3)
# We give back m tasks in a giveback phase, and each iteration we assign one task,
# so for m agents we should have a giveback probability < 1/m
auctioneer = Auctioneer(agents, tasks, 0.25) #1 / ( * len(agents)))
schedule = auctioneer.optimize_schedule()
Agent.print_schedule(X, Y, auctioneer, schedule)
def main():
agents, tasks = generate_painter(X, Y, 3)
# nFeas = The number of feasible 1-moves
# Given a schedule, we can move tasks equal to the number of agents.
# Using the result by Osman, we set alpha = n * Nfeas, gamma = n
alpha = math.factorial(X * Y)
# The number of iterations is n
annealer = Annealer(agents, tasks, 2500, 100, 0.1, 1)
schedule = annealer.optimize_schedule()
Agent.print_schedule(X, Y, annealer, schedule)
def one_X(X):
Y = X
agents, tasks = generate_painter(X, Y, parameters["num_agents"])
# https://stackoverflow.com/questions/5442910/python-multiprocessing-pool-map-for-multiple-arguments
results = []
temps = eval(parameters["temps"])
for t in temps:
results += [one_temp(t, agents, tasks)]
# https://stackoverflow.com/questions/12974474/how-to-unzip-a-list-of-tuples-into-individual-lists
means, min_cis, max_cis, mean_times = zip(*results)
with open("anneal_data/{}.csv".format(X), "w") as f:
for i in range(len(temps)):
f.write("{},{},{},{},{}\n".format(temps[i], means[i], min_cis[i],
max_cis[i], mean_times[i]))
def one_X(X):
Y = X
agents, tasks = generate_painter(X, Y, parameters["num_agents"])
# The probability of running a giveback phase should not be greater than
# 1 / |Agents|. Otherwise, it will take a VERY long time to finish, and
# probabilistically might never finish at all!
max_pr = 1 / (len(agents) + parameters["pr_epsilon"])
# Over some range of max_pr
splits = range(0, parameters["num_splits"] + 1)
# https://stackoverflow.com/questions/5442910/python-multiprocessing-pool-map-for-multiple-arguments
results = []
for s in splits:
#print(s)
results += [one_split(max_pr, s, agents, tasks)]
# https://stackoverflow.com/questions/12974474/how-to-unzip-a-list-of-tuples-into-individual-lists
means, min_cis, max_cis, mean_times = zip(*results)
with open("data/{}.csv".format(X), "w") as f:
for i in range(len(splits)):
f.write("{},{},{},{},{}\n".format(splits[i], means[i], min_cis[i], max_cis[i], mean_times[i]))
import numpy as np
import scipy.stats as st
from Base.problem_generator import generate_painter
import sys
#X_arg = int(sys.argv[1])
# The X-Y size of our grid of tasks
X = 12
NUM_TESTS = 10
NUM_SPLITS = 20
agents, tasks = generate_painter(X, X, 3)
def main():
# The probability of running a giveback phase should not be greater than
# 1 / |Agents|. Otherwise, it will take a VERY long time to finish, and
# probabilistically might never finish at all!
# Create a randomly generated schedule and judge it
s = random_schedule(agents, tasks)
auctioneer = Auctioneer(None, None, None)
results = []
for i in range(100):
results += [
auctioneer.assignment_probability(random_schedule(agents, tasks))