def generate_children(self, number_of_children, number_of_mutations, info, lookup, magnets, ref_trajectories, real_bfield=None):
# first age, as we are now creating children
self.age_bcell()
children = []
# Generate the IDfiled for the parent, as we need to calculate it fully here.
original_bfield = None
if real_bfield == None:
original_bfield, calculated_fitness = fg.calculate_cached_trajectory_fitness(info, lookup, magnets, self.genome, ref_trajectories)
fitness_error = abs(self.fitness - calculated_fitness)
logging.debug("Estimated fitness to real fitness error %2.10e"%(fitness_error))
self.fitness = calculated_fitness
else :
original_bfield = real_bfield
logging.debug("Using real bfield")
original_magnets = fg.generate_per_magnet_array(info, self.genome, magnets)
for i in range(number_of_children):
maglist = copy.deepcopy(self.genome)
maglist.mutate(number_of_mutations)
new_magnets = fg.generate_per_magnet_array(info, maglist, magnets)
update = fg.compare_magnet_arrays(original_magnets, new_magnets, lookup)
child = ID_BCell()
child.mutations = number_of_mutations
child.genome = maglist
updated_bfield = np.array(original_bfield)
for beam in update.keys() :
if update[beam].size != 0:
updated_bfield = updated_bfield - update[beam]
child.fitness = fg.calculate_trajectory_fitness_from_array(updated_bfield, info, ref_trajectories)
#child.create(info, lookup, magnets, maglist, ref_total_id_field)
children.append(child)
logging.debug("Child created with fitness : %f" % (child.fitness))
return children
def generate_children(self,
number_of_children,
number_of_mutations,
info,
lookup,
magnets,
ref_trajectories,
real_bfield=None):
# first age, as we are now creating children
self.age_bcell()
children = []
# Generate the IDfiled for the parent, as we need to calculate it fully here.
original_bfield = None
if real_bfield == None:
original_bfield, calculated_fitness = fg.calculate_cached_trajectory_fitness(
info, lookup, magnets, self.genome, ref_trajectories)
fitness_error = abs(self.fitness - calculated_fitness)
logging.debug("Estimated fitness to real fitness error %2.10e" %
(fitness_error))
self.fitness = calculated_fitness
else:
original_bfield = real_bfield
logging.debug("Using real bfield")
original_magnets = fg.generate_per_magnet_array(
info, self.genome, magnets)
for i in range(number_of_children):
maglist = copy.deepcopy(self.genome)
available_magnets = {
key: range(len(magnets.magnet_sets[key]))
for key in magnets.magnet_sets.keys()
}
maglist.mutate(number_of_mutations, available_magnets)
new_magnets = fg.generate_per_magnet_array(info, maglist, magnets)
update = fg.compare_magnet_arrays(original_magnets, new_magnets,
lookup)
child = ID_BCell()
child.mutations = number_of_mutations
child.genome = maglist
updated_bfield = np.array(original_bfield)
for beam in update.keys():
if update[beam].size != 0:
updated_bfield = updated_bfield - update[beam]
child.fitness = fg.calculate_trajectory_fitness_from_array(
updated_bfield, info, ref_trajectories)
#child.create(info, lookup, magnets, maglist, ref_total_id_field)
children.append(child)
logging.debug("Child created with fitness : %f" % (child.fitness))
return children
def create(self, info, lookup, magnets, maglist, ref_trajectories, number_of_changes, original_bfield):
self.maglist = maglist
self.changes = number_of_changes
self.create_genome(number_of_changes)
maglist = copy.deepcopy(self.maglist)
maglist.mutate_from_list(self.genome)
new_magnets = fg.generate_per_magnet_array(info, maglist, magnets)
original_magnets = fg.generate_per_magnet_array(info, self.maglist, magnets)
update = fg.compare_magnet_arrays(original_magnets, new_magnets, lookup)
updated_bfield = np.array(original_bfield)
for beam in update.keys() :
if update[beam].size != 0:
updated_bfield = updated_bfield - update[beam]
self.fitness = fg.calculate_trajectory_fitness_from_array(updated_bfield, info, ref_trajectories)
def create(self, info, lookup, magnets, maglist, ref_trajectories,
number_of_changes, original_bfield):
self.maglist = maglist
self.changes = number_of_changes
self.create_genome(number_of_changes)
maglist = copy.deepcopy(self.maglist)
maglist.mutate_from_list(self.genome)
new_magnets = fg.generate_per_magnet_array(info, maglist, magnets)
original_magnets = fg.generate_per_magnet_array(
info, self.maglist, magnets)
update = fg.compare_magnet_arrays(original_magnets, new_magnets,
lookup)
updated_bfield = np.array(original_bfield)
for beam in update.keys():
if update[beam].size != 0:
updated_bfield = updated_bfield - update[beam]
self.fitness = fg.calculate_trajectory_fitness_from_array(
updated_bfield, info, ref_trajectories)
def saveh5(path, best, genome, info, mags, real_bfield):
new_magnets = fg.generate_per_magnet_array(info, best.genome, mags)
original_magnets = fg.generate_per_magnet_array(info, genome.genome, mags)
update = fg.compare_magnet_arrays(original_magnets, new_magnets, lookup)
updated_bfield = np.array(real_bfield)
for beam in update.keys() :
if update[beam].size != 0:
updated_bfield = updated_bfield - update[beam]
outfile = os.path.join(path, genome.uid+'-'+best.uid+'.h5')
logging.debug("filename is %s" % (outfile))
f = h5py.File(outfile, 'w')
total_id_field = real_bfield
f.create_dataset('id_Bfield_original', data=total_id_field)
trajectory_information=mt.calculate_phase_error(info, total_id_field)
f.create_dataset('id_phase_error_original', data = trajectory_information[0])
f.create_dataset('id_trajectory_original', data = trajectory_information[1])
total_id_field = updated_bfield
f.create_dataset('id_Bfield_shimmed', data=total_id_field)
trajectory_information=mt.calculate_phase_error(info, total_id_field)
f.create_dataset('id_phase_error_shimmed', data = trajectory_information[0])
f.create_dataset('id_trajectory_shimmed', data = trajectory_information[1])
ref_mags=generate_reference_magnets(mags)
total_id_field = generate_id_field(info, best.genome, ref_mags, lookup)
f.create_dataset('id_Bfield_perfect', data=total_id_field)
trajectory_information=mt.calculate_phase_error(info, total_id_field)
f.create_dataset('id_phase_error_perfect', data = trajectory_information[0])
f.create_dataset('id_trajectory_perfect', data = trajectory_information[1])
f.close()
def saveh5(path, best, genome, info, mags, real_bfield, lookup):
new_magnets = fg.generate_per_magnet_array(info, best.genome, mags)
original_magnets = fg.generate_per_magnet_array(info, genome.genome, mags)
update = fg.compare_magnet_arrays(original_magnets, new_magnets, lookup)
updated_bfield = np.array(real_bfield)
for beam in update.keys():
if update[beam].size != 0:
updated_bfield = updated_bfield - update[beam]
outfile = os.path.join(path, genome.uid + '-' + best.uid + '.h5')
logging.debug("filename is %s" % (outfile))
f = h5py.File(outfile, 'w')
total_id_field = real_bfield
f.create_dataset('id_Bfield_original', data=total_id_field)
trajectory_information = mt.calculate_phase_error(info, total_id_field)
f.create_dataset('id_phase_error_original', data=trajectory_information[0])
f.create_dataset('id_trajectory_original', data=trajectory_information[1])
total_id_field = updated_bfield
f.create_dataset('id_Bfield_shimmed', data=total_id_field)
trajectory_information = mt.calculate_phase_error(info, total_id_field)
f.create_dataset('id_phase_error_shimmed', data=trajectory_information[0])
f.create_dataset('id_trajectory_shimmed', data=trajectory_information[1])
ref_mags = generate_reference_magnets(mags)
total_id_field = generate_id_field(info, best.genome, ref_mags, lookup)
f.create_dataset('id_Bfield_perfect', data=total_id_field)
trajectory_information = mt.calculate_phase_error(info, total_id_field)
f.create_dataset('id_phase_error_perfect', data=trajectory_information[0])
f.create_dataset('id_trajectory_perfect', data=trajectory_information[1])
f.close()
# send the best member of the population for sorting
trans = []
for i in range(size):
trans.append(children[0])
allpop = MPI.COMM_WORLD.alltoall(trans)
allpop.sort(key=lambda x: x.fitness)
best = allpop[0]
if rank == 0:
best.save(args[0])
new_magnets = fg.generate_per_magnet_array(info, best.genome, mags)
original_magnets = fg.generate_per_magnet_array(info, genome.genome, mags)
update = fg.compare_magnet_arrays(original_magnets, new_magnets, lookup)
updated_bfield = np.array(real_bfield)
for beam in update.keys() :
if update[beam].size != 0:
updated_bfield = updated_bfield - update[beam]
outfile = os.path.join(args[0], genome.uid+'-'+best.uid+'.h5')
logging.debug("filename is %s" % (outfile))
f = h5py.File(outfile, 'w')
total_id_field = real_bfield
f.create_dataset('id_Bfield_original', data=total_id_field)
# send the best member of the population for sorting
trans = []
for i in range(size):
trans.append(children[0])
allpop = MPI.COMM_WORLD.alltoall(trans)
allpop.sort(key=lambda x: x.fitness)
best = allpop[0]
if rank == 0:
best.save(args[0])
new_magnets = fg.generate_per_magnet_array(info, best.genome, mags)
original_magnets = fg.generate_per_magnet_array(info, genome.genome, mags)
update = fg.compare_magnet_arrays(original_magnets, new_magnets, lookup)
updated_bfield = np.array(real_bfield)
for beam in update.keys():
if update[beam].size != 0:
updated_bfield = updated_bfield - update[beam]
outfile = os.path.join(args[0], genome.uid + '-' + best.uid + '.h5')
logging.debug("filename is %s" % (outfile))
f = h5py.File(outfile, 'w')
total_id_field = real_bfield
f.create_dataset('id_Bfield_original', data=total_id_field)
def generate_children(self,
number_of_children,
number_of_mutations,
info,
lookup,
magnets,
ref_trajectories,
real_bfield=None):
# first age, as we are now creating children
self.age_bcell()
children = []
# Generate the IDfiled for the parent, as we need to calculate it fully here.
original_bfield = None
original_fitness = None
if real_bfield is None:
original_bfield, original_fitness = fg.calculate_cached_trajectory_fitness(
info, lookup, magnets, self.genome, ref_trajectories)
fitness_error = abs(self.fitness - original_fitness)
logging.debug("Estimated fitness to real fitness error %2.10e" %
(fitness_error))
else:
original_bfield = real_bfield
original_fitness = fg.calculate_trajectory_fitness_from_array(
original_bfield, info, ref_trajectories)
fitness_error = abs(self.fitness - original_fitness)
logging.debug("Using real bfield")
original_magnets = fg.generate_per_magnet_array(
info, self.maglist, magnets)
available = magnets.availability()
maglist = copy.deepcopy(self.maglist)
mutation_list = self.create_mutant(number_of_mutations, available)
maglist.mutate_from_list(mutation_list)
new_magnets = fg.generate_per_magnet_array(info, maglist, magnets)
update = fg.compare_magnet_arrays(original_magnets, new_magnets,
lookup)
updated_bfield = np.array(original_bfield)
for beam in update.keys():
if update[beam].size != 0:
updated_bfield = updated_bfield - update[beam]
self.fitness = fg.calculate_trajectory_fitness_from_array(
updated_bfield, info, ref_trajectories)
for i in range(number_of_children):
maglist = copy.deepcopy(self.maglist)
mutation_list = self.create_mutant(number_of_mutations, available)
maglist.mutate_from_list(mutation_list)
new_magnets = fg.generate_per_magnet_array(info, maglist, magnets)
update = fg.compare_magnet_arrays(original_magnets, new_magnets,
lookup)
child = ID_Shim_BCell()
child.mutations = number_of_mutations
child.genome = mutation_list
child.maglist = self.maglist
updated_bfield = np.array(original_bfield)
for beam in update.keys():
if update[beam].size != 0:
updated_bfield = updated_bfield - update[beam]
child.fitness = fg.calculate_trajectory_fitness_from_array(
updated_bfield, info, ref_trajectories)
children.append(child)
logging.debug("Child created with fitness : %2.10e" %
(child.fitness))
return children