def loglike_func(hyper,local, dat_fixm, dat_varm): sigma_like_R0, sigma_like_R1 = split_group(hyper, local) # fix mass Rob0 = dat_fixm[:,2] Rerr0 = dat_fixm[:,3] fM0 = piece_linear(hyper, M0, 0.5*np.ones_like(M0), n_pop) L0 = 0. - 0.5* np.sum( (Rob0-fM0)**2./(Rerr0**2.+sigma_like_R0**2.) ) \ - 0.5* np.sum( np.log(Rerr0**2.+sigma_like_R0**2.) ) # variable mass Mob1 = dat_varm[:,0] Merr1 = dat_varm[:,1] Rob1 = dat_varm[:,2] Rerr1 = dat_varm[:,3] Mth1 = local + 0. fM1 = piece_linear(hyper, Mth1, 0.5*np.ones_like(Mth1), n_pop) L1 = 0. - 0.5* np.sum( (Mob1-Mth1)**2./Merr1**2. ) \ - 0.5* np.sum( (Rob1-fM1)**2./(Rerr1**2.+sigma_like_R1**2.) ) \ - 0.5* np.sum( np.log(Rerr1**2.+sigma_like_R1**2.) ) L = L0 + L1 return L
def inverse_local(local_prob, hyper): # R(0,i) for fix mass prob_R0 = local_prob[0:n_fixm] R0 = piece_linear(hyper, M0, prob_R0) # M(1,i) for variable mass prob_M1 = local_prob[n_fixm:n_fixm+n_varm] M1 = uniform.ppf(prob_M1, -4., 10.) # R(1,i) for varibable mass prob_R1 = local_prob[n_fixm+n_varm:] R1 = piece_linear(hyper, M1, prob_R1) local = np.hstack((R0, M1, R1)) return local
def Mpost2R(mass, unit='Earth', classify='No'):
"""
Forecast the Radius distribution given the mass distribution.
Parameters
---------------
mass: one dimensional array
The mass distribution.
unit: string (optional)
Unit of the mass.
Options are 'Earth' and 'Jupiter'. Default is 'Earth'.
classify: string (optional)
If you want the object to be classifed.
Options are 'Yes' and 'No'. Default is 'No'.
Result will be printed, not returned.
Returns
---------------
radius: one dimensional array
Predicted radius distribution in the input unit.
"""
# mass input
mass = np.array(mass)
assert len(mass.shape) == 1, "Input mass must be 1-D."
# unit input
if unit == 'Earth':
pass
elif unit == 'Jupiter':
mass = mass * mearth2mjup
else:
print(
"Input unit must be 'Earth' or 'Jupiter'. Using 'Earth' as default."
)
# mass range
if np.min(mass) < 3e-4 or np.max(mass) > 3e5:
print('Mass range out of model expectation. Returning None.')
return None
## convert to radius
sample_size = len(mass)
logm = np.log10(mass)
prob = np.random.random(sample_size)
logr = np.ones_like(logm)
hyper_ind = np.random.randint(low=0,
high=np.shape(all_hyper)[0],
size=sample_size)
hyper = all_hyper[hyper_ind, :]
if classify == 'Yes':
classification(logm, hyper[:, -3:])
for i in range(sample_size):
logr[i] = piece_linear(hyper[i], logm[i], prob[i])
radius_sample = 10.**logr
## convert to right unit
if unit == 'Jupiter':
radius = radius_sample / rearth2rjup
else:
radius = radius_sample
return radius
# find_best.py with spatial_median.txt and find trans_1up/down trans_1up = trans_best + np.array([0.1245372 , 0.0534476 , 0.04035559]) trans_1down = trans_best - np.array([0.14317994, 0.06079727, 0.03514506 ]) # convert trans best mearth2mjup = 317.828 mearth2msun = 333060.4 m_trans = 10.**trans_best m1 = m_trans[0] m2 = m_trans[1] / mearth2mjup m3 = m_trans[2] / mearth2msun # find the corresponding R at transition points rad_trans = 10.** piece_linear( best_hyper, trans_best, prob_R = 0.5*np.ones(3) ) print 'radius at transition (RE)/(RJ)', rad_trans, rad_trans/11.21 ### plot plt.clf() figscale=1.5 fig = plt.figure(figsize=(4*figscale,3*figscale)) ax = plt.gca() # fit line m_max = np.log10(np.max(star[:,0])) m_min = np.log10(np.min(dwarfplanet[:,0])) m_sample = np.linspace( m_min, m_max, 1000 ) r_sample = piece_linear(best_hyper, m_sample, prob_R = 0.5*np.ones_like(m_sample))
def Mpost2R(mass, unit='Earth', classify='No'):
"""
Forecast the Radius distribution given the mass distribution.
Parameters
---------------
mass: one dimensional array
The mass distribution.
unit: string (optional)
Unit of the mass.
Options are 'Earth' and 'Jupiter'. Default is 'Earth'.
classify: string (optional)
If you want the object to be classifed.
Options are 'Yes' and 'No'. Default is 'No'.
Result will be printed, not returned.
Returns
---------------
radius: one dimensional array
Predicted radius distribution in the input unit.
"""
# mass input
mass = np.array(mass)
assert len(mass.shape) == 1, "Input mass must be 1-D."
# unit input
if unit == 'Earth':
pass
elif unit == 'Jupiter':
mass = mass * mearth2mjup
else:
print("Input unit must be 'Earth' or 'Jupiter'. "
"Using 'Earth' as default.")
# mass range
if np.min(mass) < 3e-4 or np.max(mass) > 3e5:
print('Mass range out of model expectation. Returning None.')
return None
## convert to radius
sample_size = len(mass)
logm = np.log10(mass)
prob = np.random.random(sample_size)
logr = np.ones_like(logm)
hyper_ind = np.random.randint(low=0, high=np.shape(all_hyper)[0],
size=sample_size)
hyper = all_hyper[hyper_ind,:]
if classify == 'Yes':
classification(logm, hyper[:,-3:])
for i in range(sample_size):
logr[i] = piece_linear(hyper[i], logm[i], prob[i])
radius_sample = 10.** logr
## convert to right unit
if unit == 'Jupiter':
radius = radius_sample / rearth2rjup
else:
radius = radius_sample
return radius
def plt_linear(infile, outfile, datadir=""):
pp = PdfPages(outfile)
### data
hyper = np.loadtxt(infile + "hyper.out")
loglike = np.loadtxt(infile + "loglike.out")
repeat = np.loadtxt(infile + "repeat.out")
### split data
c0 = hyper[:, 0]
slope = hyper[:, 1:5]
sigma = hyper[:, 5:9]
trans = hyper[:, 9:12]
### plot
plt.clf()
row = 2
col = 4
f, ((a00, a01, a02, a03), (a10, a11, a12, a13)) = plt.subplots(row, col, figsize=(col * 5, row * 5))
ax = ((a00, a01, a02, a03), (a10, a11, a12, a13))
# repeat
ax[0][0].plot(repeat)
ax[0][0].set_yscale("log")
ax[0][0].set_xlabel("repeat")
# loglike
ax[0][1].plot(loglike)
ax[0][1].set_xlabel("L")
# loglike
ax[0][2].plot(range(len(loglike) / 2, len(loglike)), loglike[len(loglike) / 2 :])
ax[0][1].set_xlabel("L")
# over plot
dat = convert_data(np.loadtxt(datadir + "PlanetGroup.txt"))
ax[0][3].errorbar(dat[:, 0], dat[:, 2], xerr=dat[:, 1], yerr=dat[:, 3], fmt=".")
best_ind = np.argmax(loglike)
hyper_best = hyper[best_ind, :]
trans_best = hyper_best[-3:]
print 10.0 ** trans_best
m_sample = np.linspace(np.min(dat[:, 0]), np.max(dat[:, 0]), 1000)
r_sample = piece_linear(hyper_best, m_sample, prob_R=0.5 * np.ones_like(m_sample))
# r_upper = piece_linear(hyper_best, m_sample, prob_R = 0.84 * np.ones_like(m_sample))
# r_lower = piece_linear(hyper_best, m_sample, prob_R = 0.16 * np.ones_like(m_sample))
r_upper = piece_linear_complex(hyper_best, m_sample, prob_R=0.84 * np.ones_like(m_sample))
r_lower = piece_linear_complex(hyper_best, m_sample, prob_R=0.16 * np.ones_like(m_sample))
ax[0][3].plot(m_sample, r_sample, "r-")
ax[0][3].fill_between(m_sample, r_lower, r_upper, color="grey", alpha=0.2)
r_trans = piece_linear(hyper_best, trans_best, prob_R=0.5 * np.ones_like(trans_best))
ax[0][3].plot(trans_best, r_trans, "rx")
ax[0][3].set_xlabel(r"log10(M [M$_\oplus$])")
ax[0][3].set_ylabel(r"log10(R [R$_\oplus$])")
# C
ax[1][0].plot(c0)
ax[1][0].set_xlabel("c0")
ax[1][0].set_ylim([-1, 1])
# slope
for i in range(4):
ax[1][1].plot(slope[:, i])
ax[1][1].set_xlabel("slope")
# sigma
for i in range(4):
ax[1][2].plot(sigma[:, i])
ax[1][2].set_yscale("log")
ax[1][2].set_xlabel("sigma")
ax[1][2].set_ylim([1e-3, 1e0])
# transition
for i in range(3):
ax[1][3].plot(trans[:, i])
ax[1][3].set_xlabel("transition")
ax[1][3].set_ylim([-4, 6])
pp.savefig()
pp.close()
return None
r2m_m = np.log10(Rpost2M(10.**r2m_r))
r2m_r2 = np.log10(Mpost2R(10.**r2m_m))
### mass histogram
nbin=10
ax1.hist(m2r_m2, fc='r', ec='r', alpha=0.3, bins=nbin)
ax1.hist(m2r_m,fc='b', bins=nbin, label=['M -> R -> M'])
ax1.hist(r2m_m,fc='g', bins=nbin, label=['R -> M'])
ax1.legend(bbox_to_anchor=(1.05, 1), loc=2, borderaxespad=0.)
### model
from func import piece_linear, piece_linear_complex
best_hyper = np.loadtxt('spatial_median.txt', delimiter=',')
m_sample = np.linspace( -3.9, 5.5, 1000 )
r_sample = piece_linear(best_hyper, m_sample, prob_R = 0.5*np.ones_like(m_sample))
r_upper = piece_linear_complex(best_hyper, m_sample, prob_R = 0.84 * np.ones_like(m_sample))
r_lower = piece_linear_complex(best_hyper, m_sample, prob_R = 0.16 * np.ones_like(m_sample))
r_2upper = piece_linear_complex(best_hyper, m_sample, prob_R = 0.975 * np.ones_like(m_sample))
r_2lower = piece_linear_complex(best_hyper, m_sample, prob_R = 0.025 * np.ones_like(m_sample))
ax2.plot(m_sample, r_sample, 'r-')
ax2.fill_between(m_sample, r_lower, r_upper, color='grey', alpha=0.6)
ax2.fill_between(m_sample, r_2lower, r_2upper, color='grey', alpha=0.4)
ax2.set_xlabel(r'$\rm log_{10}\ M/M_\oplus$')
ax2.set_ylabel(r'$\rm log_{10}\ R/R_\oplus$')
### radius histogram
ax3.hist(m2r_r,fc='b', orientation='horizontal', bins=nbin)
ax3.hist(r2m_r,fc='g', orientation='horizontal', bins=nbin)
