class Benchmarker(object):
'Main class for benchmarker.'
def __init__(self):
self.launcher = Launcher()
def go(self):
'''
Launches the benchmarker.
1.loads problem
1. Clears queue and results.
2. create new job
3. Executes the queue and retrieves the best result and plots it
4. New variable and back to [1].
'''
print 'Go benchmarker go!'
#base is a job based on the results of the optimizer
base = {
'evaluate': 'eval_simple',
'mutate': ('mutshuf', {'indpb': 0.01}),
'mate': 'cxSCX',
'population': 201,
'generations': 248,
'indices': 'path_creator',
'cxpb': 0.80,
'mutpb': 0.2,
}
problems = ['belgiumtour.tsp', 'xqf131.tsp', 'bcl380.tsp', 'xql622.tsp']
for problem in problems:
#1. load problem
box = Box('data/TSPBenchmark')
data = box.get(problem)
self.launcher.load_data(data)
#2. clear
self.launcher.clear()
#3. create job
print 'Attacking problem: ', problem
job = base.copy()
job['name'] = 'bmark-{}'.format(problem)
#4. launch job
best = self.launcher.launch_queue([job], plot_all=True)
# plot solution
analysis.plot(best['best'][0], data)
print 'done!'
raw_input()
def __init__(self):
self.launcher = Launcher()
class Optimizer(object):
'Main class for parameter optimization.'
def __init__(self):
self.launcher = Launcher()
def go(self):
'''
Launches the super uber optimizer.
1. Clears queue and results.
2. Queues new results for variable range of interest
3. Executing the queue and retrieving the best result
4. New variable and back to [2].
'''
print 'Go go go !'
# 0. Loading data
box = Box('data/TSPBenchmark')
data = box.get('xqf131.tsp')
self.launcher.load_data(data)
# 1. clear queue and results
self.launcher.clear()
# 2. Create dictionnaries for range
base = {
'evaluate': 'eval_simple',
'mutate': ('mutshuf', {'indpb': 0.01}),
'mate': 'cxSCX',
'population': 201,
'generations': 248,
'indices': 'path_creator',
'cxpb': 0.80,
'mutpb': 0.2,
}
print 'Optimizing population'
jobs = []
for pop in range(1, 1000, 50):
job = base.copy()
job['name'] = 'pop-{}'.format(pop)
job['population'] = pop
job['generations'] = 10000 / pop
jobs.append(job)
best = self.launcher.launch_queue(jobs, plot_all=False)['set']
del best['_id']
print 'Optimizing cxpb'
jobs = []
for cxpb in xrange(1, 100, 5):
job = best.copy()
job['name'] = '{}-{}'.format(job['name'], cxpb / 100.)
job['cxpb'] = cxpb / 100.
jobs.append(job)
best = self.launcher.launch_queue(jobs, plot_all=False)['set']
del best['_id']
print 'Optimizing indpb'
jobs = []
for mut in range(1, 100, 5):
job = best.copy()
job['name'] = '{}-{}'.format(job['name'], mut / 100.)
job['mutate'] = (job['mutate'][0], {'indpb': mut / 100.})
jobs.append(job)
best = self.launcher.launch_queue(jobs, plot_all=False)['set']
del best['_id']
print 'Optimizing mutpb'
jobs = []
for mutpb in xrange(1, 100, 5):
job = best.copy()
job['name'] = '{}-{}'.format(job['name'], mutpb / 100.)
job['mutpb'] = mutpb / 100.
jobs.append(job)
best = self.launcher.launch_queue(jobs, plot_all=False)['set']
del best['_id']
print 'Optimizing crossover operator'
crossovers = ['cxPMX', 'cxSCX', 'cxERX', 'cxOX', 'cxHeuristic']
jobs = []
for cx_operator in crossovers:
job = best.copy()
job['name'] = '{}-{}'.format(job['name'], cx_operator)
job['mate'] = cx_operator
if cx_operator == 'cxHeuristic':
job['evaluate'] = 'eval_adjacent'
job['mutate'] = ('mutshuf_adj', {'indpb': best['mutate'][1]})
job['indices'] = 'adj_creator'
else:
job['evaluate'] = 'eval_simple'
job['mutate'] = ('mutshuf', {'indpb': best['mutate'][1]})
job['indices'] = 'path_creator'
jobs.append(job)
best = self.launcher.launch_queue(jobs, plot_all=True)['set']
del best['_id']
print 'Optimizing mutation operator'
jobs = []
if best['mate'] == 'cxHeuristic':
mutations = ['mutshuf_adj', 'invert_mut_adj', 'insert_mut_adj', 'simple_inv_adj']
else:
mutations = ['mutshuf', 'invert_mut', 'insert_mut', 'simple_inv']
for mut_operator in mutations:
job = best.copy()
job['name'] = '{}-{}'.format(job['name'], mut_operator)
job['mutate'] = (mut_operator, {'indpb': best['mutate'][1]})
jobs.append(job)
best = self.launcher.launch_queue(jobs, plot_all=True)['set']
del best['_id']
print 'runned'
raw_input()
def __init__(self):
self.launcher = Launcher()
class RComparator(object):
'Main class for representation comparison.'
def __init__(self):
self.launcher = Launcher()
def go(self):
# '''
# Launches the super uber optimizer.
# 1. Clears queue and results.
# 2. Queues new results for variable range of interest
# 3. Executing the queue and retrieving the best result
# 4. New variable and back to [2].
# '''
print 'Go go power rangers!'
# 0. Loading data
box = Box('data/TSPBenchmark')
data = box.get('xqf131.tsp')
self.launcher.load_data(data)
# 1. clear queue and results
self.launcher.clear()
# 2. Create base dictionary
base = {
'evaluate': 'eval_simple',
'mutate': ('mutshuf', {'indpb': 0.01}),
'mate': 'cxSCX',
'population': 201,
'generations': 2,
'indices': 'path_creator',
'cxpb': 0.80,
'mutpb': 0.2,
}
adj_mutations = ['mutshuf_adj', 'invert_mut_adj', 'insert_mut_adj', 'simple_inv_adj']
path_mutations = ['mutshuf', 'invert_mut', 'insert_mut', 'simple_inv']
crossovers = ['cxPMX', 'cxESCX', 'cxSCX', 'cxERX', 'cxOX', 'cxHeuristic']
for mut_operator in path_mutations:
print 'Computing comparison for ', mut_operator
jobs = []
for crossover in crossovers:
job = base.copy()
job['name'] = '{}-{}'.format(crossover, mut_operator)
job['mate'] = crossover
if crossover == 'cxHeuristic':
job['evaluate'] = 'eval_adjacent'
job['indices'] = 'adj_creator'
# we set the adjacent mutation relying on the fact that adj mut and path mut
# follow the same ordering
index_curr_mut = path_mutations.index(mut_operator)
job['mutate'] = (adj_mutations[index_curr_mut], {'indpb': job['mutate'][1]['indpb']})
else:
job['evaluate'] = 'eval_simple'
job['mutate'] = (mut_operator, {'indpb': job['mutate'][1]['indpb']})
job['indices'] = 'path_creator'
jobs.append(job)
self.launcher.launch_queue(jobs, plot_all=True)['set']
print 'runned'
raw_input()
class Optimizer(object):
'Main class for parameter optimization.'
def __init__(self):
self.launcher = Launcher()
def go(self):
'''
Launches the super uber optimizer.
1. Clears queue and results.
2. Queues new results for variable range of interest
3. Executing the queue and retrieving the best result
4. New variable and back to [2].
'''
print 'Go go go !'
# 0. Loading data
box = Box('data/TSPBenchmark')
data = box.get('xqf131.tsp')
self.launcher.load_data(data)
# 1. clear queue and results
self.launcher.clear()
# 2. Create dictionnaries for range
base = {
'evaluate': 'eval_simple',
'mutate': ('mutshuf', {
'indpb': 0.01
}),
'mate': 'cxSCX',
'population': 201,
'generations': 248,
'indices': 'path_creator',
'cxpb': 0.80,
'mutpb': 0.2,
}
print 'Optimizing population'
jobs = []
for pop in range(1, 1000, 50):
job = base.copy()
job['name'] = 'pop-{}'.format(pop)
job['population'] = pop
job['generations'] = 10000 / pop
jobs.append(job)
best = self.launcher.launch_queue(jobs, plot_all=False)['set']
del best['_id']
print 'Optimizing cxpb'
jobs = []
for cxpb in xrange(1, 100, 5):
job = best.copy()
job['name'] = '{}-{}'.format(job['name'], cxpb / 100.)
job['cxpb'] = cxpb / 100.
jobs.append(job)
best = self.launcher.launch_queue(jobs, plot_all=False)['set']
del best['_id']
print 'Optimizing indpb'
jobs = []
for mut in range(1, 100, 5):
job = best.copy()
job['name'] = '{}-{}'.format(job['name'], mut / 100.)
job['mutate'] = (job['mutate'][0], {'indpb': mut / 100.})
jobs.append(job)
best = self.launcher.launch_queue(jobs, plot_all=False)['set']
del best['_id']
print 'Optimizing mutpb'
jobs = []
for mutpb in xrange(1, 100, 5):
job = best.copy()
job['name'] = '{}-{}'.format(job['name'], mutpb / 100.)
job['mutpb'] = mutpb / 100.
jobs.append(job)
best = self.launcher.launch_queue(jobs, plot_all=False)['set']
del best['_id']
print 'Optimizing crossover operator'
crossovers = ['cxPMX', 'cxSCX', 'cxERX', 'cxOX', 'cxHeuristic']
jobs = []
for cx_operator in crossovers:
job = best.copy()
job['name'] = '{}-{}'.format(job['name'], cx_operator)
job['mate'] = cx_operator
if cx_operator == 'cxHeuristic':
job['evaluate'] = 'eval_adjacent'
job['mutate'] = ('mutshuf_adj', {'indpb': best['mutate'][1]})
job['indices'] = 'adj_creator'
else:
job['evaluate'] = 'eval_simple'
job['mutate'] = ('mutshuf', {'indpb': best['mutate'][1]})
job['indices'] = 'path_creator'
jobs.append(job)
best = self.launcher.launch_queue(jobs, plot_all=True)['set']
del best['_id']
print 'Optimizing mutation operator'
jobs = []
if best['mate'] == 'cxHeuristic':
mutations = [
'mutshuf_adj', 'invert_mut_adj', 'insert_mut_adj',
'simple_inv_adj'
]
else:
mutations = ['mutshuf', 'invert_mut', 'insert_mut', 'simple_inv']
for mut_operator in mutations:
job = best.copy()
job['name'] = '{}-{}'.format(job['name'], mut_operator)
job['mutate'] = (mut_operator, {'indpb': best['mutate'][1]})
jobs.append(job)
best = self.launcher.launch_queue(jobs, plot_all=True)['set']
del best['_id']
print 'runned'
raw_input()
class RComparator(object):
'Main class for representation comparison.'
def __init__(self):
self.launcher = Launcher()
def go(self):
# '''
# Launches the super uber optimizer.
# 1. Clears queue and results.
# 2. Queues new results for variable range of interest
# 3. Executing the queue and retrieving the best result
# 4. New variable and back to [2].
# '''
print 'Go go power rangers!'
# 0. Loading data
box = Box('data/TSPBenchmark')
data = box.get('xqf131.tsp')
self.launcher.load_data(data)
# 1. clear queue and results
self.launcher.clear()
# 2. Create base dictionary
base = {
'evaluate': 'eval_simple',
'mutate': ('mutshuf', {
'indpb': 0.01
}),
'mate': 'cxSCX',
'population': 201,
'generations': 2,
'indices': 'path_creator',
'cxpb': 0.80,
'mutpb': 0.2,
}
adj_mutations = [
'mutshuf_adj', 'invert_mut_adj', 'insert_mut_adj', 'simple_inv_adj'
]
path_mutations = ['mutshuf', 'invert_mut', 'insert_mut', 'simple_inv']
crossovers = [
'cxPMX', 'cxESCX', 'cxSCX', 'cxERX', 'cxOX', 'cxHeuristic'
]
for mut_operator in path_mutations:
print 'Computing comparison for ', mut_operator
jobs = []
for crossover in crossovers:
job = base.copy()
job['name'] = '{}-{}'.format(crossover, mut_operator)
job['mate'] = crossover
if crossover == 'cxHeuristic':
job['evaluate'] = 'eval_adjacent'
job['indices'] = 'adj_creator'
# we set the adjacent mutation relying on the fact that adj mut and path mut
# follow the same ordering
index_curr_mut = path_mutations.index(mut_operator)
job['mutate'] = (adj_mutations[index_curr_mut], {
'indpb': job['mutate'][1]['indpb']
})
else:
job['evaluate'] = 'eval_simple'
job['mutate'] = (mut_operator, {
'indpb': job['mutate'][1]['indpb']
})
job['indices'] = 'path_creator'
jobs.append(job)
self.launcher.launch_queue(jobs, plot_all=True)['set']
print 'runned'
raw_input()