/
george_tracks.py
137 lines (99 loc) · 3.72 KB
/
george_tracks.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
import matplotlib.pyplot as plt
import numpy as np
from grids import DartmouthPMS, PISA, Baraffe15, Seiss
# grid = DartmouthPMS(age_range=[1, 200], mass_range=[0.5, 2.5])
# grid = PISA(age_range=[1, 100], mass_range=[0.5, 2.0])
grid = Baraffe15(age_range=[1, 100], mass_range=[0.5, 1.4])
# grid = Seiss(age_range=[1, 100], mass_range=[0.5, 2.0])
grid.load()
# Try computing a GP
import george
from george.kernels import ExpSquaredKernel
# Set up the Gaussian process.
A_temp, A_radius, tau_age, tau_mass = np.array([3057., 1.0, 3.3, 0.19])
temp_kernel = A_temp**2 * ExpSquaredKernel(metric=[tau_age**2, tau_mass**2], ndim=2)
radius_kernel = A_radius**2 * ExpSquaredKernel(metric=[tau_age**2, tau_mass**2], ndim=2)
N = len(grid.points)
temp_gp = george.GP(temp_kernel)
radius_gp = george.GP(radius_kernel)
# print(temp_gp.kernel.pars)
# temp_gp.kernel[:] = np.log(np.array([A_temp, tau_age, tau_mass, nugget_temp])**2)
#
# print(temp_gp.kernel.pars)
# plt.imshow(temp_gp.get_matrix(grid.points), interpolation="none", origin="upper")
# plt.colorbar()
# plt.savefig("matrix.png")
# Pre-compute the factorization of the matrix.
temp_gp.compute(grid.points, yerr=50.*np.ones(N))
radius_gp.compute(grid.points, yerr=0.3*np.ones(N))
#
# # # Compute the log likelihood.
# print(temp_gp.lnlikelihood(grid.temps))
# print(radius_gp.lnlikelihood(grid.radii))
# import sys
# sys.exit()
fig, ax = plt.subplots(nrows=1, figsize=(8,8))
blue = True
for mass in np.unique(grid.masses):
track = np.array([np.linspace(2., 20., num=50), mass * np.ones(50)]).T
mu_T, cov = temp_gp.predict(grid.temps, track)
# Ts = temp_gp.sample_conditional(grid.temps, track, 5)
# std_T = np.sqrt(np.diag(cov))
mu_R, cov = radius_gp.predict(grid.radii, track)
# std_R = np.sqrt(np.diag(cov))
# Rs = radius_gp.sample_conditional(grid.radii, track, 5)
# if blue:
# col = "b."
# else:
# col = "k."
# blue = not blue
#
# for T,R in zip(Ts, Rs):
# ax.plot(T, R, col, ms=1)
ax.plot(mu_T, mu_R)
# Label the tau0, M points
masses = np.unique(grid.masses)
for mass in masses:
# Find all T, R that have this mass
ind = (grid.masses == mass)
tt = grid.temps[ind]
rr = grid.radii[ind]
ax.plot(tt, rr, "g-")
ax.plot(tt, rr, "go")
# ax.annotate("{:.1f}".format(mass), (tt[0], rr[0]), size=5)
# ax.plot(grid.temps, grid.radii, "go")
ax.set_xlim(8000, 3000)
fig.savefig("TR.png")
def lnprob(p):
A_temp, A_radius, tau_age, tau_mass = p
if np.any(p <= 0):
return -np.inf
# Setting the "vector", so we need to use the natural log: http://dan.iel.fm/george/current/user/kernels/#implementation
temp_gp.kernel[:] = np.log(np.array([A_temp, tau_age, tau_mass])**2)
radius_gp.kernel[:] = np.log(np.array([A_radius, tau_age, tau_mass])**2)
lnp = temp_gp.lnlikelihood(grid.temps, quiet=True) + radius_gp.lnlikelihood(grid.radii, quiet=True)
# print("P:", p, "lnp:", lnp)
return lnp
def optimize():
from emcee import EnsembleSampler
import multiprocessing as mp
ndim = 4
nwalkers = 4 * ndim
p0 = np.array([np.random.uniform(1000, 5000, nwalkers),
np.random.uniform(0.1, 1.0, nwalkers),
np.random.uniform(2, 12, nwalkers),
np.random.uniform(0.1, 1.5, nwalkers)]).T
sampler = EnsembleSampler(nwalkers, ndim, lnprob, threads=mp.cpu_count())
pos, prob, state = sampler.run_mcmc(p0, 1000)
sampler.reset()
print("Burned in")
# actual run
pos, prob, state = sampler.run_mcmc(pos, 1000)
# Save the last position of the walkers
np.save("walkers_emcee.npy", pos)
np.save("eparams_emcee.npy", sampler.flatchain)
def main():
# optimize()
pass
if __name__=="__main__":
main()