def sample(self):
"""
Method to pick the sample satisfying the likelihood constraint using uniform sampling
Returns
-------
new : object
The evolved sample
number : int
Number of likelihood calculations after sampling
"""
new = Source()
x_l, x_u = self.getPrior_X()
y_l, y_u = self.getPrior_Y()
r_l, r_u = self.getPrior_R()
a_l, a_u = self.getPrior_A()
while(True):
new.X = np.random.uniform(x_l,x_u)
new.Y = np.random.uniform(y_l,y_u)
new.A = np.random.uniform(a_l,a_u)
new.R = np.random.uniform(r_l,r_u)
new.logL = self.log_likelihood(new)
self.number+=1
if(new.logL > self.LC):
break
return new, self.number
def sample(self):
"""
Method to pick the sample satisfying the likelihood constraint using uniform sampling
Returns
-------
new : object
The evolved sample
number : int
Number of likelihood calculations after sampling
"""
new = Source()
x_l, x_u = self.getPrior_X()
y_l, y_u = self.getPrior_Y()
r_l, r_u = self.getPrior_R()
a_l, a_u = self.getPrior_A()
while (True):
new.X = np.random.uniform(x_l, x_u)
new.Y = np.random.uniform(y_l, y_u)
new.A = np.random.uniform(a_l, a_u)
new.R = np.random.uniform(r_l, r_u)
new.logL = self.log_likelihood(new)
self.number += 1
if (new.logL > self.LC):
break
return new, self.number
def sample_source(self):
"""
Sampling the object from prior distribution.
Returns
-------
src : object
The source object with X,Y,A,R sampled from their prior distribution and log likelihood calculated.
"""
src = Source()
src.X = random.uniform(0.0, self.x_upper)
src.Y = random.uniform(0.0, self.y_upper)
src.A = random.uniform(self.amplitude_lower, self.amplitude_upper)
src.R = random.uniform(self.R_lower, self.R_upper)
src.logL = self.log_likelihood(src)
return src
def sample(self):
"""
Sampling from the built ellipsoids.
Returns
-------
clust : object
The sampled point satisfying the likelihood constraint
number : int
The number of likelihood calculations until this
"""
trial = Source()
clust = Source()
z = int(np.random.uniform(0, len(self.ellipsoid_set)))
points = None
try:
points = self.ellipsoid_set[z].sample(n_points=50)
except IndexError:
raise Exception(
"Adjust clustering parameters or number of active samples")
max_likelihood = self.LC
count = 0
r_l, r_u = self.getPrior_R()
a_l, a_u = self.getPrior_A()
while count < 50:
trial.X = points[count][0]
trial.Y = points[count][1]
trial.A = np.random.uniform(a_l, a_u)
trial.R = np.random.uniform(r_l, r_u)
trial.logL = self.log_likelihood(trial)
self.number += 1
if (trial.logL > max_likelihood):
clust.__dict__ = trial.__dict__.copy()
max_likelihood = trial.logL
break
count += 1
return clust, self.number
def sample(self):
"""
Method to pick the sample satisfying the likelihood constraint using metropolis sampling
Returns
-------
metro : object
The evolved sample
number : int
Number of likelihood calculations until now
"""
metro = Source()
metro.__dict__ = self.source.__dict__.copy()
start = Source()
start.__dict__ = self.source.__dict__.copy()
new = Source()
self.number+=1
count = 0
hit = 0
miss = 0
x_l, x_u = self.getPrior_X()
y_l, y_u = self.getPrior_Y()
r_l, r_u = self.getPrior_R()
a_l, a_u = self.getPrior_A()
stepnormalize = self.step/x_u
stepX = self.step
stepY = stepnormalize*(y_u-y_l)
stepA = stepnormalize*(a_u - a_l)
stepR = stepnormalize*(r_u-r_l)
bord = 1
while(count<20):
while bord==1:
bord = 0
new.X = metro.X + stepX * (2.*np.random.uniform(0, 1) - 1.);
new.Y = metro.Y + stepY * (2.*np.random.uniform(0, 1) - 1.);
new.A = metro.A + stepA * (2.*np.random.uniform(0, 1) - 1.);
new.R = metro.R + stepR * (2.*np.random.uniform(0, 1) - 1.);
if(new.X > x_u or new.X < x_l): bord = 1;
if(new.Y > y_u or new.Y < y_l): bord = 1;
if(new.A > a_u or new.A < a_l): bord = 1;
if(new.R > r_u or new.R < r_l): bord = 1;
new.logL = self.log_likelihood(new)
self.number+=1
if(new.logL > self.LC):
metro.__dict__ = new.__dict__.copy()
hit+=1
else:
miss+=1
if( hit > miss ): self.step *= exp(1.0 / hit);
if( hit < miss ): self.step /= exp(1.0 / miss);
stepnormalize = self.step/x_u
stepX = self.step
stepY = stepnormalize*(y_u-y_l)
stepA = stepnormalize*(a_u - a_l)
stepR = stepnormalize*(r_u-r_l)
count+=1
bord=1
return metro, self.number
def sample(self):
"""
Method to pick the sample satisfying the likelihood constraint using metropolis sampling
Returns
-------
metro : object
The evolved sample
number : int
Number of likelihood calculations until now
"""
metro = Source()
metro.__dict__ = self.source.__dict__.copy()
start = Source()
start.__dict__ = self.source.__dict__.copy()
new = Source()
self.number += 1
count = 0
hit = 0
miss = 0
x_l, x_u = self.getPrior_X()
y_l, y_u = self.getPrior_Y()
r_l, r_u = self.getPrior_R()
a_l, a_u = self.getPrior_A()
stepnormalize = self.step / x_u
stepX = self.step
stepY = stepnormalize * (y_u - y_l)
stepA = stepnormalize * (a_u - a_l)
stepR = stepnormalize * (r_u - r_l)
bord = 1
while (count < 20):
while bord == 1:
bord = 0
new.X = metro.X + stepX * (2. * np.random.uniform(0, 1) - 1.)
new.Y = metro.Y + stepY * (2. * np.random.uniform(0, 1) - 1.)
new.A = metro.A + stepA * (2. * np.random.uniform(0, 1) - 1.)
new.R = metro.R + stepR * (2. * np.random.uniform(0, 1) - 1.)
if (new.X > x_u or new.X < x_l): bord = 1
if (new.Y > y_u or new.Y < y_l): bord = 1
if (new.A > a_u or new.A < a_l): bord = 1
if (new.R > r_u or new.R < r_l): bord = 1
new.logL = self.log_likelihood(new)
self.number += 1
if (new.logL > self.LC):
metro.__dict__ = new.__dict__.copy()
hit += 1
else:
miss += 1
if (hit > miss): self.step *= exp(1.0 / hit)
if (hit < miss): self.step /= exp(1.0 / miss)
stepnormalize = self.step / x_u
stepX = self.step
stepY = stepnormalize * (y_u - y_l)
stepA = stepnormalize * (a_u - a_l)
stepR = stepnormalize * (r_u - r_l)
count += 1
bord = 1
return metro, self.number
