def nfi_phase(self, prev_position, prev_fitness):
new_position_list = []
xichma_v = 1
xichma_u = ((gamma(1 + 1.5) * np.sin(np.pi * 1.5 / 2)) / (gamma(
(1 + 1.5) / 2) * 1.5 * 2**((1.5 - 1) / 2)))**(1.0 / 1.5)
self.levy_b = (np.random.normal(0, xichma_u**2)) / (np.sqrt(
np.abs(np.random.normal(0, xichma_v**2)))**(1.0 / 1.5))
# NFi phase
Pb = np.random.uniform()
Pfi = np.random.uniform()
self.alpha = 0.01
for i in range(self.pop_size):
# Calculate neutron vector Nei by Eq. (2)
# Random 1 more index to select neutron
temp1 = list(set(range(self.pop_size)) - {i})
i1 = np.random.choice(temp1, replace=False)
Nei = (prev_position[i] + prev_position[i1]) / 2
# Update population of fission products according to Eq.(3), (6) or (9);
if np.random.uniform() <= Pfi:
# Update based on Eq. 3
if np.random.uniform() <= Pb:
xichma1 = (loge(self.current_epoch + 1) * 1.0 /
(self.current_epoch + 1)) * np.abs(
np.subtract(prev_position[i],
self.g_best_pos))
gauss = np.array([
np.random.normal(self.g_best_pos[j], xichma1[j])
for j in range(self.problem_size)
])
Xi = (gauss + np.random.uniform() * self.g_best_pos -
np.round(np.random.rand() + 1) * Nei)
# Update based on Eq. 6
else:
i2 = np.random.choice(temp1, replace=False)
xichma2 = (loge(self.current_epoch + 1) * 1.0 /
(self.current_epoch + 1)) * np.abs(
np.subtract(prev_position[i2],
self.g_best_pos))
gauss = np.array([
np.random.normal(prev_position[i][j], xichma2[j])
for j in range(self.problem_size)
])
Xi = (gauss + np.random.uniform() * self.g_best_pos -
np.round(np.random.rand() + 2) * Nei)
# Update based on Eq. 9
else:
i3 = np.random.choice(temp1, replace=False)
xichma2 = (loge(self.current_epoch + 1) * 1.0 /
(self.current_epoch + 1)) * np.abs(
np.subtract(prev_position[i3], self.g_best_pos))
Xi = np.array([
np.random.normal(prev_position[i][j], xichma2[j])
for j in range(self.problem_size)
])
new_position_list.append(Xi)
return new_position_list
vodata['_r'] == n.min(vodata['_r']))[0]]['Seq'][0]
src_logL = -99
src_logLs = [src_logL]
else:
try: #Use the flux maximum likelihood method to choose the optimal match d
src_logLs = []
for seq in seqs:
spec = n.sort(vodata[n.where(vodata['Seq'] == seq)[0]],
order='nu').squeeze()
src_logL = -(log10(src.jys)-P[seq](log10(src.mfreq)))**2/\
(2*((log10(src.e_S_nu+src.jys)-log10(src.jys))**2+e_P[seq](src.mfreq)**2)) +\
-n.max(spec['_r'])**2/(2*sres**2)
src_logLs.append(src_logL)
log.debug('calculated %d likelihoods' % len(src_logLs))
src_logLs = n.array(src_logLs)
src_logLs -= loge(sum(exp(src_logLs)))
src_logL = n.max(src_logLs)
seq = seqs[n.where(src_logLs == src_logL)[0]]
except (AttributeError):
# print "This catalog does not have flux errors"
src_logL = -99
#if not enough flux info, use the nearest position only
seq = vodata[n.where(
vodata['_r'] == n.min(vodata['_r']))[0]]['Seq'][0]
#put in the data for the matched source
for seq in seqs:
VO.data['logL_cat'][n.where(
VO.data['Seq'] == seq)[0]] = src_logLs
VO.data['S_sf'][n.where(
VO.data['Seq'] == seq)[0]] = 10**P[seq](log10(src.mfreq))
def train(self):
pop = [
self.create_solution(minmax=self.ID_MIN_PROB)
for _ in range(self.pop_size)
]
g_best = self.get_global_best_solution(pop, self.ID_FIT,
self.ID_MIN_PROB)
for epoch in range(self.epoch):
xichma_v = 1
xichma_u = ((gamma(1 + 1.5) * sin(pi * 1.5 / 2)) / (gamma(
(1 + 1.5) / 2) * 1.5 * 2**((1.5 - 1) / 2)))**(1.0 / 1.5)
levy_b = (normal(0, xichma_u**2)) / (sqrt(
abs(normal(0, xichma_v**2)))**(1.0 / 1.5))
# NFi phase
Pb = uniform()
Pfi = uniform()
freq = 0.05
alpha = 0.01
for i in range(self.pop_size):
## Calculate neutron vector Nei by Eq. (2)
## Random 1 more index to select neutron
temp1 = list(set(range(0, self.pop_size)) - {i})
i1 = choice(temp1, replace=False)
Nei = (pop[i][self.ID_POS] + pop[i1][self.ID_POS]) / 2
## Update population of fission products according to Eq.(3), (6) or (9);
if uniform() <= Pfi:
### Update based on Eq. 3
if uniform() <= Pb:
xichma1 = (loge(epoch + 1) * 1.0 / (epoch + 1)) * abs(
subtract(pop[i][self.ID_POS], g_best[self.ID_POS]))
gauss = array([
normal(g_best[self.ID_POS][j], xichma1[j])
for j in range(self.problem_size)
])
Xi = gauss + uniform() * g_best[self.ID_POS] - round(
rand() + 1) * Nei
### Update based on Eq. 6
else:
i2 = choice(temp1, replace=False)
xichma2 = (loge(epoch + 1) * 1.0 / (epoch + 1)) * abs(
subtract(pop[i2][self.ID_POS],
g_best[self.ID_POS]))
gauss = array([
normal(pop[i][self.ID_POS][j], xichma2[j])
for j in range(self.problem_size)
])
Xi = gauss + uniform() * g_best[self.ID_POS] - round(
rand() + 2) * Nei
## Update based on Eq. 9
else:
i3 = choice(temp1, replace=False)
xichma2 = (loge(epoch + 1) * 1.0 / (epoch + 1)) * abs(
subtract(pop[i3][self.ID_POS], g_best[self.ID_POS]))
Xi = array([
normal(pop[i][self.ID_POS][j], xichma2[j])
for j in range(self.problem_size)
])
## Check the boundary and evaluate the fitness function
Xi = self.amend_position_random_faster(Xi)
fit = self.get_fitness_position(Xi, self.ID_MIN_PROB)
if fit < pop[i][self.ID_FIT]:
pop[i] = [Xi, fit]
if fit < g_best[self.ID_FIT]:
g_best = [Xi, fit]
# NFu phase
## Ionization stage
## Calculate the Pa through Eq. (10);
ranked_pop = rankdata(
[pop[i][self.ID_FIT] for i in range(self.pop_size)])
for i in range(self.pop_size):
X_ion = deepcopy(pop[i][self.ID_POS])
if (ranked_pop[i] * 1.0 / self.pop_size) < uniform():
i1, i2 = choice(list(set(range(0, self.pop_size)) - {i}),
2,
replace=False)
for j in range(self.problem_size):
#### Levy flight strategy is described as Eq. 18
if pop[i2][self.ID_POS][j] == pop[i][self.ID_POS][j]:
X_ion[j] = pop[i][self.ID_POS][
j] + alpha * levy_b * (pop[i][self.ID_POS][j] -
g_best[self.ID_POS][j])
#### If not, based on Eq. 11, 12
else:
if uniform() <= 0.5:
X_ion[j] = pop[i1][self.ID_POS][j] + uniform(
) * (pop[i2][self.ID_POS][j] -
pop[i][self.ID_POS][j])
else:
X_ion[j] = pop[i1][self.ID_POS][j] - uniform(
) * (pop[i2][self.ID_POS][j] -
pop[i][self.ID_POS][j])
else: #### Levy flight strategy is described as Eq. 21
X_worst = self.get_global_best_solution(
pop, self.ID_FIT, self.ID_MAX_PROB)
for j in range(self.problem_size):
##### Based on Eq. 21
if X_worst[self.ID_POS][j] == g_best[self.ID_POS][j]:
X_ion[j] = pop[i][self.ID_POS][
j] + alpha * levy_b * (self.ub[j] - self.lb[j])
##### Based on Eq. 13
else:
X_ion[j] = pop[i][self.ID_POS][j] + round(uniform(
)) * uniform() * (X_worst[self.ID_POS][j] -
g_best[self.ID_POS][j])
## Check the boundary and evaluate the fitness function for X_ion
X_ion = self.amend_position_random_faster(X_ion)
fit = self.get_fitness_position(X_ion, self.ID_MIN_PROB)
if fit < pop[i][self.ID_FIT]:
pop[i] = [X_ion, fit]
if fit < g_best[self.ID_FIT]:
g_best = [X_ion, fit]
## Fusion Stage
### all ions obtained from ionization are ranked based on (14) - Calculate the Pc through Eq. (14)
ranked_pop = rankdata(
[pop[i][self.ID_FIT] for i in range(self.pop_size)])
for i in range(self.pop_size):
i1, i2 = choice(list(set(range(0, self.pop_size)) - {i}),
2,
replace=False)
#### Generate fusion nucleus
if (ranked_pop[i] * 1.0 / self.pop_size) < uniform():
t1 = uniform() * (pop[i1][self.ID_POS] -
g_best[self.ID_POS])
t2 = uniform() * (pop[i2][self.ID_POS] -
g_best[self.ID_POS])
temp2 = pop[i1][self.ID_POS] - pop[i2][self.ID_POS]
X_fu = pop[i][
self.ID_POS] + t1 + t2 - exp(-norm(temp2)) * temp2
#### Else
else:
##### Based on Eq. 22
check_equal = (
pop[i1][self.ID_POS] == pop[i2][self.ID_POS])
if check_equal.all():
X_fu = pop[i][self.ID_POS] + alpha * levy_b * (
pop[i][self.ID_POS] - g_best[self.ID_POS])
##### Based on Eq. 16, 17
else:
if uniform() > 0.5:
X_fu = pop[i][self.ID_POS] - 0.5 * (
sin(2 * pi * freq * epoch + pi) *
(self.epoch - epoch) / self.epoch + 1
) * (pop[i1][self.ID_POS] - pop[i2][self.ID_POS])
else:
X_fu = pop[i][self.ID_POS] - 0.5 * (
sin(2 * pi * freq * epoch + pi) * epoch /
self.epoch + 1) * (pop[i1][self.ID_POS] -
pop[i2][self.ID_POS])
X_fu = self.amend_position_random_faster(X_fu)
fit = self.get_fitness_position(X_fu, self.ID_MIN_PROB)
if fit < pop[i][self.ID_FIT]:
pop[i] = [X_fu, fit]
if fit < g_best[self.ID_FIT]:
g_best = [X_fu, fit]
self.loss_train.append(g_best[self.ID_FIT])
if self.verbose:
print(">Epoch: {}, Best fit: {}".format(
epoch + 1, g_best[self.ID_FIT]))
self.solution = g_best
return g_best[self.ID_POS], g_best[self.ID_FIT], self.loss_train
if len(seqs)<2:#if we only found one source, the IDing is easy
seq = vodata[n.where(vodata['_r']==n.min(vodata['_r']))[0]]['Seq'][0]
src_logL = -99
src_logLs = [src_logL]
else:
try: #Use the flux maximum likelihood method to choose the optimal match d
src_logLs = []
for seq in seqs:
spec = n.sort(vodata[n.where(vodata['Seq']==seq)[0]],order='nu').squeeze()
src_logL = -(log10(src.jys)-P[seq](log10(src.mfreq)))**2/\
(2*((log10(src.e_S_nu+src.jys)-log10(src.jys))**2+e_P[seq](src.mfreq)**2)) +\
-n.max(spec['_r'])**2/(2*sres**2)
src_logLs.append(src_logL)
log.debug('calculated %d likelihoods'%len(src_logLs))
src_logLs =n.array(src_logLs)
src_logLs -= loge(sum(exp(src_logLs)))
src_logL = n.max(src_logLs)
seq = seqs[n.where(src_logLs==src_logL)[0]]
except(AttributeError):
# print "This catalog does not have flux errors"
src_logL = -99
#if not enough flux info, use the nearest position only
seq = vodata[n.where(vodata['_r']==n.min(vodata['_r']))[0]]['Seq'][0]
#put in the data for the matched source
for seq in seqs:
VO.data['logL_cat'][n.where(VO.data['Seq']==seq)[0]] = src_logLs
VO.data['S_sf'][n.where(VO.data['Seq']==seq)[0]]=10**P[seq](log10(src.mfreq))
VO.data['e_S_sf'][n.where(VO.data['Seq']==seq)[0]]=\
10**(P[seq](log10(src.mfreq))+res[seq]) -\
10**P[seq](log10(src.mfreq))