Repository navigation
Expand file tree
/
Copy pathloglikelihood.py
More file actions
133 lines (94 loc) · 3.87 KB
/
Copy pathloglikelihood.py
File metadata and controls
133 lines (94 loc) · 3.87 KB
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
import jax
import jax.numpy as jnp
import numpy as np
import scipy
from model import jax_stream_model, jax_stream_orbit
BAD_VAL = -1e100
@jax.jit
def loglikelihood_stream(p, dict_data):
logM, Rs, q, dirx, diry, dirz, logm, rs, x0, z0, vx0, vy0, vz0, time, alpha = p
dirz = jnp.abs(dirz)
x0 = jnp.abs(x0)
z0 = jnp.abs(z0)
vy0 = jnp.abs(vy0)
r_data = dict_data['r_data']
w_data = dict_data['w_data']
r_err = dict_data['r_err']
w_err = dict_data['w_err']
y0 = 0.
_, _, _, _, r_meds, w_meds, _, _, _ = jax_stream_model(logM, Rs, q, dirx, diry, dirz, logm, rs, x0, y0, z0, vx0, vy0, vz0, time, alpha, tail=0, min_count=101)
mask = ~jnp.isnan(r_data)
# Count how many predictions are bad
nan_mask = jnp.isnan(jnp.where(mask, r_meds, 0.0))
n_bad = jnp.sum(nan_mask)
def all_nan_case(_):
return BAD_VAL * r_data.shape[0] #-jnp.inf #
def some_good_case(_):
def good_fit_case(_):
res = ((r_meds - r_data) / r_err) ** 2
return -0.5 * jnp.nansum(res)
def bad_fit_case(_):
return BAD_VAL * n_bad #-jnp.inf #
return jax.lax.cond(n_bad == 0, good_fit_case, bad_fit_case, operand=None)
logl = jax.lax.cond(jnp.all(jnp.isnan(r_meds)), all_nan_case, some_good_case, operand=None)
return logl
@jax.jit
def loglikelihood_orbit(p, dict_data):
logM, Rs, q, dirx, diry, dirz, x0, z0, vx0, vy0, vz0, time, alpha = p
dirz = jnp.abs(dirz)
x0 = jnp.abs(x0)
z0 = jnp.abs(z0)
vy0 = jnp.abs(vy0)
r_data = dict_data['r_data']
w_data = dict_data['w_data']
r_err = dict_data['r_err']
w_err = dict_data['w_err']
y0 = 0.
_, _, _, _, r_meds, w_meds, _, _, _ = jax_stream_orbit(logM, Rs, q, dirx, diry, dirz, x0, y0, z0, vx0, vy0, vz0, time, alpha)
mask = ~jnp.isnan(r_data)
# Count how many predictions are bad
nan_mask = jnp.isnan(jnp.where(mask, r_meds, 0.0))
n_bad = jnp.sum(nan_mask)
def good_fit_case(_):
res = ((r_meds - r_data) / r_err) ** 2
return -0.5 * jnp.nansum(res)
def bad_fit_case(_):
return -jnp.inf #BAD_VAL * r_data.shape[0]
logl = jax.lax.cond(n_bad == 0, good_fit_case, bad_fit_case, operand=None)
return logl
@jax.jit
def loglikelihood_data(p, r_data, r_err):
logM, Rs, q, dirx, diry, dirz, logm, rs, x0, z0, vx0, vy0, vz0, time, alpha, sig = p
# r_data = dict_data['r_data']
# w_data = dict_data['w_data']
# r_err = dict_data['r_err']
# w_err = dict_data['w_err']
y0 = 0.
_, _, _, _, r_meds, w_meds, _, _, _ = jax_stream_model(logM, Rs, q, dirx, diry, dirz, logm, rs, x0, y0, z0, vx0, vy0, vz0, time, alpha, tail=0, min_count=101)
mask = ~jnp.isnan(r_data)
# Count how many predictions are bad
nan_mask = jnp.isnan(jnp.where(mask, r_meds, 0.0))
n_bad = jnp.sum(nan_mask)
def all_nan_case(_):
return -jnp.inf #BAD_VAL * r_data.shape[0] #-jnp.inf #
def some_good_case(_):
def good_fit_case(_):
res = (r_meds - r_data)**2 / (r_err**2 + sig**2) + jnp.log(r_err**2 + sig**2)
return -0.5 * jnp.nansum(res)
def bad_fit_case(_):
return -jnp.inf #BAD_VAL * n_bad #-jnp.inf #
return jax.lax.cond(n_bad == 0, good_fit_case, bad_fit_case, operand=None)
logl = jax.lax.cond(jnp.all(jnp.isnan(r_meds)), all_nan_case, some_good_case, operand=None)
return logl
def wrapper_loglikelihood_data(p, r_data, r_err):
logM, Rs, dirx, diry, dirz, logm, rs, x0, z0, vx0, vy0, vz0, t0, a0, sig0 = p
r = np.sqrt(dirx**2 + diry**2 + dirz**2)
q = np.exp(-r**2/2) * (np.sqrt(np.pi) * np.exp(r**2/2) * scipy.special.erf(r/np.sqrt(2)) - np.sqrt(2)*r)/np.sqrt(np.pi)
q += 0.5
p1 = jnp.array([
logM, Rs, q, dirx, diry, dirz,
logm, rs,
x0, z0, vx0, vy0, vz0,
t0, a0, sig0
])
return loglikelihood_data(p1, r_data, r_err)