def run_point(sx,sy,sI,dx,dy,geometry,cross_sections):
x = sx
y = sy
I = sI
dA = 0.005806 # m^2
#detector inputs: location (x,y), epsilon = 0.62, area, dwell = 5.0 secs
detectors = [gefry3.Detector((dx,dy), 0.62, dA, 5) for i in range(2)]
#--- Setup Domain ---#
print("1")
with open(geometry) as f:
geo = pickle.load(f) #geo is a list of shapely polygons
print("2")
buildings = [gefry3.Solid(s.exterior.coords) for s in geo]
bbox = [(0, 0), (250, 0), (250, 178), (0, 178), (0, 0)]
domain = gefry3.Domain(bbox, buildings)
sigmas = np.loadtxt(cross_sections)
materials = [gefry3.Material(1,s) for s in sigmas] #this is a bit of a hack, these are the building cross sections
#Interstitial material is just air. Don't think this was in gefry2.
interstitial_material = gefry3.Material(1,0) #cross section of the air
# I don't think this is correct, but let's assume the air doesn't attenuate rads
#--- Setup Source ---#
true_source = gefry3.Source((158,98),3.214e9) #not really used for anything
#--- Initalize Radiation Model ---#
P = gefry3.SimpleProblem(domain, interstitial_material, materials, true_source, detectors)
return P((x,y),I)
def abc(prior,background,observations,detectors,geometry,cross_sections):
obs = np.loadtxt(observations)
n_obs = obs.shape[0]
n_sen = obs.shape[1]
r_d = np.loadtxt(detectors)
#--- Setup Detectors ---#
dA = 0.005806 # m^2
#detector inputs: location (x,y), epsilon = 0.62, area, dwell = 5.0 secs
detectors = [gefry3.Detector(i, 0.62, dA, 5) for i in r_d]
#--- Setup Domain ---#
with open(geometry) as f:
geo = pickle.load(f) #geo is a list of shapely polygons
buildings = [gefry3.Solid(s.exterior.coords) for s in geo]
bbox = [(0, 0), (250, 0), (250, 178), (0, 178), (0, 0)]
domain = gefry3.Domain(bbox, buildings)
sigmas = np.loadtxt(cross_sections)
materials = [gefry3.Material(1,s) for s in sigmas] #this is a bit of a hack, these are the building cross sections
#Interstitial material is just air. Don't think this was in gefry2.
interstitial_material = gefry3.Material(1,0) #cross section of the air
# I don't think this is correct, but let's assume the air doesn't attenuate rads
#--- Setup Source ---#
true_source = gefry3.Source((158,98),3.214e9) #not really used for anything
#--- Initalize Radiation Model ---#
P = gefry3.SimpleProblem(domain, interstitial_material, materials, true_source, detectors)
#--- ABC run ---#
N = prior.shape[0]
result = np.zeros((N,n_sen))
for i in range(N):
if ((i % 1000) == 0): print(i/N)
result[i,] = np.random.poisson(np.rint(P((prior[i,0],prior[i,1]),prior[i,2])+background))
return result
buildings = [gefry3.Solid(s.exterior.coords) for s in geo]
bbox = [(0, 0), (250, 0), (250, 178), (0, 178), (0, 0)]
domain = gefry3.Domain(bbox, buildings) #geo is a list of shapely polygons
sigmas = np.loadtxt('cross_sec.dat') #these go in the interstitial materials component?
materials = [gefry3.Material(1,s) for s in sigmas] #this is a bit of a hack, these are the building cross sections
interstitial_material = gefry3.Material(1,0) #cross section of the air the gammas are moving through (but we don't have this?)
# I don't think this is correct, but let's try it
#--- Setup Source ---#
true_source = gefry3.Source((158,98),3.214e9)
#--- Initalize Radiation Model ---#
P = gefry3.SimpleProblem(domain, interstitial_material, materials, true_source, detectors)
mean_response = P((85,85),3.214e9) + background
# file_name = 'radiation_chain%d.txt' % (init_values[4])
# np.savetxt(file_name,chain, delimiter=',')
#
obs = []
for i in range(n_obs):
obs.append(np.random.poisson(mean_response))
obs_filename = file_prefix + "_obs"
np.savetxt(obs_filename,obs, delimiter=' ')
#
# def radiation_model(x, y, I, detectors, n_obs):
# sensor_response = P((x, y), I) # .astype(np.float64)
def radiation_model_sampler(init_values):
np.random.seed(init_values[3])
x = mc.Uniform('x', lower=0, upper=246.615, value=init_values[0])
y = mc.Uniform('y', lower=0, upper=176.333, value=init_values[1])
I = mc.Uniform('I', lower=5e8, upper=5e10, value=init_values[2])
n_obs = 10 #num observations per detector
obs = np.loadtxt('obs.dat')
#--- Setup Detectors ---#
dA = 0.005806 # m^2
r_d = np.loadtxt('det.dat') #jk different order of the stuff
detectors = [
gefry3.Detector(i, 0.62, dA, 5) for i in r_d
] # for each detector location, dwell time = 5.0, epsilon = 0.62 (efficiency)
#--- Setup Domain ---#
with open('./revised_geo.pkl') as f:
geo = pickle.load(f)
buildings = [gefry3.Solid(s.exterior.coords) for s in geo]
bbox = [(0, 0), (250, 0), (250, 178), (0, 178), (0, 0)]
domain = gefry3.Domain(bbox, buildings) #geo is a list of shapely polygons
sigmas = np.loadtxt(
'cross_sec.dat') #these go in the interstitial materials component?
materials = [
gefry3.Material(1, s) for s in sigmas
] #this is a bit of a hack, these are the building cross sections
interstitial_material = gefry3.Material(
1, 0
) #cross section of the air the gammas are moving through (but we don't have this?)
# I don't think this is correct, but let's try it
#domain = gefry2.Domain(geo, sigma)
#--- Setup Source ---#
true_source = gefry3.Source((158, 98), 3.214e9)
#--- Initalize Radiation Model ---#
#P = gefry2.Problem(domain, detectors, S, 300)
P = gefry3.SimpleProblem(domain, interstitial_material, materials,
true_source, detectors)
#print(P((158,98),3.214e9))
background_count = 300
def radiation_model(x, y, I, detectors, n_obs):
sensor_response = P((x, y), I) # .astype(np.float64)
return (sensor_response)
#Likelihood
@mc.stochastic(observed=True)
def rad_counts(value=obs, x=x, y=y, I=I):
output = np.rint(P((x, y), I)) + background_count
LogLik = 0
for i in range(0, n_obs):
LogLik = LogLik + (np.sum(obs[:, i] * np.log(output[i])) -
n_obs * output[i])
return LogLik
model = mc.MCMC([x, y, I, rad_counts])
model.use_step_method(mc.AdaptiveMetropolis, [x, y, I],
shrink_if_necessary=True)
#model.sample(iter=10**4+3000, burn=3000)
model.sample(iter=20, burn=0, verbose=0) #testing purposes
# Chains
samples_x = model.trace('x')[:]
samples_y = model.trace('y')[:]
samples_I = model.trace('I')[:]
chain = np.vstack((samples_x, samples_y, samples_I)).T
file_name = 'radiation_chain%d.txt' % (init_values[4])
np.savetxt(file_name, chain, delimiter=',')
return 0