-
Notifications
You must be signed in to change notification settings - Fork 0
/
plot_foundation_results.py
294 lines (220 loc) · 12.9 KB
/
plot_foundation_results.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
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
import matplotlib.pyplot as plt
import matplotlib.lines as mlines
from astropy.cosmology import FlatLambdaCDM
import pickle
import numpy as np
import corner
import pandas as pd
fiducial_cosmology={"H0": 73.24, "Om0": 0.28}
cosmo = FlatLambdaCDM(**fiducial_cosmology)
num_to_plot = 157
dataset = 'T21_training_set'
sn_list = pd.read_csv('data/lcs/tables/' + dataset + '.txt', comment='#', delim_whitespace=True, names=['sn', 'source', 'files'])
sn_names = list(sn_list.sn.values)
sn_names[sn_names.index("AT2016ajl")] = 'AT2016aj'
av_stat = 'median'
title_str = ""
sn_info = pd.read_csv("data/lcs/Foundation_DR1/Foundation_DR1.FITRES.TEXT", comment='#', delim_whitespace=True)
class Result:
def __init__(self, filename):
self.results_dict = np.load(filename, allow_pickle=True).item()
self.mu_samples = self.results_dict['mu'].reshape((num_to_plot, 1000,))
self.theta_samples = self.results_dict['theta'].reshape((num_to_plot, 1000,))
self.av_samples = self.results_dict['AV'].reshape((num_to_plot, 1000,))
self.samples_dict = {'mu':self.mu_samples, 'theta':self.theta_samples, 'AV':self.av_samples}
self.point_estimates = {var: np.median(self.samples_dict[var], axis=1) for var in self.samples_dict.keys()}
self.stds = {var: np.std(self.samples_dict[var], axis=1) for var in self.samples_dict.keys()}
self.variances = {var: np.var(self.samples_dict[var], axis=1) for var in self.samples_dict.keys()}
# need to re-run MCMC with new model -> this is done
# zltn_result = Result("foundation_vmap_zltn_122923.npy")
# mcmc_result = Result("foundation_vmap_mcmc_122923.npy")
# laplace_result = Result("foundation_vmap_laplace_122923.npy")
# multinormal_result = Result("foundation_vmap_multinormal_122923.npy")
zltn_result = Result("foundation_results/foundation_vmap_zltn_032624_samples.npy")
mcmc_result = Result("foundation_results/foundation_vmap_mcmc_032624_samples.npy")
laplace_result = Result("foundation_results/foundation_vmap_laplace_032624_samples.npy")
multinormal_result = Result("foundation_results/foundation_vmap_multinormal_032624_samples.npy")
zcmb_dict = dict(zip(sn_info.CID.values, sn_info.zCMB.values))
zcmbs = np.array([zcmb_dict[name] for name in sn_names])
def get_mode_from_samples(samples):
hist, bin_edges = np.histogram(samples, bins=50)
max_index = np.argmax(hist)
mode = (bin_edges[max_index] + bin_edges[max_index + 1])/2
return mode
# load Stephen's MCMC chains for comparison
# using names to make sure they're in the same order
stephen_point_estimates = {'mu':[], 'theta':[], 'AV':[]}
stephen_uncertainties = {'mu':[], 'theta':[], 'AV':[]}
for i in range(num_to_plot):
s = np.load("../dist_chains_210610_135216/" + sn_names[i] + "_chains_210610_135216.npy", allow_pickle=True).item()
for var in ['mu', 'theta', "AV"]:
stephen_point_estimates[var].append(np.median(s[var]))
stephen_uncertainties[var].append(np.std(s[var]))
for i in range(num_to_plot - 1):
if (zltn_result.point_estimates['theta'][i] - mcmc_result.point_estimates['theta'][i]).any() < -1.5:
print(i, sn_names[i])
for i in range(num_to_plot):
if np.isnan(zltn_result.point_estimates['AV'][i]):
print(i, sn_names[i])
best_ks = np.loadtxt("foundation_results/best_ks_032824.txt")
last_ks = np.loadtxt("foundation_results/last_ks_032824.txt")
best_samples = np.load("foundation_results/best_samples_032824.npz", allow_pickle = True)['arr_0']
last_samples = np.load("foundation_results/last_samples_032824.npz", allow_pickle = True)['arr_0']
best_mu_medians = np.array([np.median(best_samples[i]['Ds']) for i in range(num_to_plot)])
last_mu_medians = np.array([np.median(last_samples[i]['Ds']) for i in range(num_to_plot)])
best_mu_std = np.array([np.std(best_samples[i]['Ds']) for i in range(num_to_plot)])
last_mu_std = np.array([np.std(last_samples[i]['Ds']) for i in range(num_to_plot)])
fig, ax = plt.subplots()
t = ax.scatter(best_ks, last_ks, c = best_mu_medians - last_mu_medians, cmap='bwr')
linspace_vals = np.linspace(0.3, 1.3)
ax.plot(linspace_vals,linspace_vals, 'k')
cbar = fig.colorbar(t)
cbar.set_label('$\\Delta \\mu$')
ax.set_xlabel("k from best-loss parameters")
ax.set_ylabel("k from last-loss parameters")
plt.show()
fig.savefig("figures/bestvslastk.pdf", bbox_inches='tight')
plt.scatter(best_ks, last_ks, c = best_mu_medians - np.array([cosmo.distmod(z).value for z in zcmbs]), cmap='bwr')
linspace_vals = np.linspace(0.3, 1.3)
plt.plot(linspace_vals,linspace_vals, 'k')
cbar = plt.colorbar()
cbar.set_label('$D_{s, best} - \\mu(z)$')
plt.xlabel("k from best loss params")
plt.ylabel("k from last iteration params")
plt.show()
plt.scatter(best_ks, last_ks, c = best_mu_std - last_mu_std, cmap='bwr')
linspace_vals = np.linspace(0.3, 1.3)
plt.plot(linspace_vals,linspace_vals, 'k')
cbar = plt.colorbar()
cbar.set_label('$D_{s, best} - \\mu(z)$')
plt.xlabel("k from best loss params")
plt.ylabel("k from last iteration params")
plt.show()
plt.plot(best_ks - last_ks, best_mu_std - last_mu_std, 'o')
plt.xlabel("delta k (best - last)")
plt.ylabel("delta stdev (best - last)")
plt.show()
fig = plt.figure()
plt.plot(last_ks, last_mu_medians - mcmc_result.point_estimates['mu'], 'o')
plt.axvline(0.7, linestyle='dashed', color='k')
plt.xlabel("$\\hat{k}$", fontsize = 20)
plt.xticks(fontsize=18)
plt.yticks(fontsize=18)
plt.ylabel("MVZLTN VI $\\mu$ - MCMC $\\mu$", fontsize = 20)
fig.savefig("figures/khat_mu_comparison.pdf", bbox_inches = 'tight')
plt.show()
plt.plot(last_ks, last_mu_medians - np.array([cosmo.distmod(z).value for z in zcmbs]), 'o')
plt.xlabel("$\hat{k}$")
plt.ylabel("VI $\mu$ - $\\mu(z)$")
plt.show()
# regularized_k = (96 * last_ks + 10*0.5) / (96 + 10)
# plt.plot(regularized_k, last_mu_medians - mcmc_result.point_estimates['mu'], 'o')
# plt.xlabel("$\hat{k}$")
# plt.ylabel("VI $\mu$ - MCMC $\mu$")
# plt.show()
# VI vs MCMC subplots
alpha = 0.2
fig, ax = plt.subplots(2,3, figsize = (9,9))
mcmc_color = 'r'
vi_color = 'b'
axis_fontsize = 12
latex_version = {'mu': '$\\mu$', 'theta': '$\\theta$', 'AV': '$A_V$'}
for i, var in enumerate(['mu', 'theta', 'AV']):
ax[0][i].plot(mcmc_result.point_estimates[var], zltn_result.point_estimates[var], '.', alpha=alpha, c=vi_color, label='VI')
ax[0][i].errorbar(mcmc_result.point_estimates[var], zltn_result.point_estimates[var], yerr = zltn_result.stds[var], xerr = mcmc_result.stds[var], alpha=alpha, c=vi_color, linestyle='None')
linspace_vals = np.linspace(min(mcmc_result.point_estimates[var]), max(mcmc_result.point_estimates[var]))
ax[0][i].plot(linspace_vals, linspace_vals, c='k')
ax[0][i].set_ylabel('VI '+ latex_version[var], fontsize = axis_fontsize)
ax[1][i].plot(mcmc_result.point_estimates[var], zltn_result.point_estimates[var] - mcmc_result.point_estimates[var], '.', alpha=alpha, c = vi_color, label='VI')
ax[1][i].errorbar(mcmc_result.point_estimates[var], zltn_result.point_estimates[var] - mcmc_result.point_estimates[var], xerr = mcmc_result.stds[var], yerr = zltn_result.stds[var], alpha=alpha, c=vi_color, linestyle='None')
ax[1][i].axhline(0, color = 'k')
ax[1][i].set_ylabel('Residual ' + latex_version[var] + ' (VI - MCMC)', fontsize = axis_fontsize)
ax[1][i].set_xlabel('NumPyro MCMC ' + latex_version[var], fontsize = axis_fontsize)
# ax[2][i].plot(true_values[var], zltn_result.point_estimates[var] - mcmc_result.point_estimates[var], 'o', color='gray')
# # ax[2][i].errorbar(true_values[var], zltn_result.point_estimates[var] - mcmc_result.point_estimates[var], yerr = np.max([zltn_result.stds[var], mcmc_result.stds[var]]), c='gray', linestyle='None')
# ax[2][i].axhline(0, color = 'k')
# ax[2][i].set_ylabel('Residual ' + latex_version[var] + ' (VI - MCMC)', fontsize = axis_fontsize)
for axis in ax.flatten():
axis.tick_params(axis='x', labelsize=12)
axis.tick_params(axis='y', labelsize=12)
plt.tight_layout()
plt.show()
## Plot comparison to Stephen's MCMC chains
fig, ax = plt.subplots(2,3, figsize = (9,9))
for i, var in enumerate(['mu', 'theta', 'AV']):
ax[0][i].plot(stephen_point_estimates[var], zltn_result.point_estimates[var], '.', alpha=alpha, c=vi_color, label='VI')
ax[0][i].errorbar(stephen_point_estimates[var], zltn_result.point_estimates[var], yerr = zltn_result.stds[var], xerr = stephen_uncertainties[var], alpha=alpha, c=vi_color, linestyle='None')
linspace_vals = np.linspace(min(stephen_point_estimates[var]), max(stephen_point_estimates[var]))
ax[0][i].plot(linspace_vals, linspace_vals, c='k')
ax[0][i].set_ylabel('VI '+ latex_version[var], fontsize = axis_fontsize)
ax[1][i].plot(stephen_point_estimates[var], zltn_result.point_estimates[var] - stephen_point_estimates[var], '.', alpha=alpha, c = vi_color, label='VI')
ax[1][i].errorbar(stephen_point_estimates[var], zltn_result.point_estimates[var] - stephen_point_estimates[var], yerr = zltn_result.stds[var], xerr = stephen_uncertainties[var], alpha=alpha, c=vi_color, linestyle='None')
ax[1][i].axhline(0, color = 'k')
ax[1][i].set_ylabel('Residual ' + latex_version[var] + ' (VI - MCMC)', fontsize = axis_fontsize)
ax[1][i].set_xlabel('Stan MCMC ' + latex_version[var], fontsize = axis_fontsize)
# ax[2][i].plot(true_values[var], zltn_result.point_estimates[var] - mcmc_result.point_estimates[var], 'o', color='gray')
# # ax[2][i].errorbar(true_values[var], zltn_result.point_estimates[var] - mcmc_result.point_estimates[var], yerr = np.max([zltn_result.stds[var], mcmc_result.stds[var]]), c='gray', linestyle='None')
# ax[2][i].axhline(0, color = 'k')
# ax[2][i].set_ylabel('Residual ' + latex_version[var] + ' (VI - MCMC)', fontsize = axis_fontsize)
for axis in ax.flatten():
axis.tick_params(axis='x', labelsize=12)
axis.tick_params(axis='y', labelsize=12)
plt.tight_layout()
plt.show()
#### plot MCMC vs VI uncertainties
fig, ax = plt.subplots(1,3)
for i, var in enumerate(['mu', 'theta', 'AV']):
ax[i].plot(zltn_result.stds[var], mcmc_result.stds[var], 'o')
min_value = np.min(np.concatenate((zltn_result.stds[var], mcmc_result.stds[var])))
max_value = np.max(np.concatenate((zltn_result.stds[var], mcmc_result.stds[var])))
linspace_vals = np.linspace(min_value, max_value)
ax[i].plot(linspace_vals, linspace_vals, 'k')
ax[i].set_ylabel('MCMC ' + latex_version[var] + ' uncertainty', fontsize = axis_fontsize)
ax[i].set_xlabel('VI ' + latex_version[var] + ' uncertainty', fontsize = axis_fontsize)
plt.show()
## Plot Hubble diagram
linspace_z = np.linspace(0.01, 0.085)
cosmo_distmod_values = np.array([cosmo.distmod(z).value for z in linspace_z])
def plot_hubble_distances_and_residuals(mus, variances, z_cmbs, linspace_z = linspace_z, cosmo_distmod_values = cosmo_distmod_values):
predictions = mus
targets = np.array([cosmo.distmod(z).value for z in z_cmbs])
rmse = np.sqrt(np.mean((predictions-targets)**2))
fig, ax = plt.subplots(2,1, figsize = (6,9), sharex='col', gridspec_kw = {'hspace':0})
ax[0].errorbar(z_cmbs, mus, np.sqrt(variances), linestyle = 'None', color = 'k')
ax[0].plot(z_cmbs, mus,'o')
ax[0].plot(linspace_z, cosmo_distmod_values, color = 'k')
ax[0].set_xlabel("z", fontsize =20)
ax[0].set_ylabel("$\\mu$", fontsize = 20)
print(rmse)
plt.text(0.07, 0.8, "Foundation DR1", fontsize = 16, horizontalalignment = 'center')
plt.text(0.07, 0.7, "N = " + str(len(z_cmbs)), fontsize = 16, horizontalalignment = 'center')
plt.text(0.07, 0.6, "RMSE = " + '{:.3f}'.format(round(rmse, 3)), fontsize = 16, horizontalalignment = 'center')
sigma_pec = 150
c = 300000
print(np.mean(mus - np.array(targets)), "+/-", np.std(mus - np.array(targets)) / np.sqrt(len(z_cmbs)))
residuals = mus - np.array(targets)
print("reduced chisq:", sum(residuals**2/variances) / len(mus) - 1)
ax[1].errorbar(z_cmbs, mus - np.array(targets), np.sqrt(variances), linestyle = 'None', color = 'k')
ax[1].plot(z_cmbs, mus - np.array(targets), 'o')
sigma_envelope = np.array([(5 / (z * np.log(10))) * (sigma_pec / c) for z in linspace_z])
ax[1].plot(linspace_z, sigma_envelope, marker = 'None', linestyle = 'dashed', color = 'k')
ax[1].plot(linspace_z, -sigma_envelope, marker = 'None', linestyle = 'dashed', color = 'k')
ax[1].axhline(0., color = 'k')
ax[1].set_xlabel("z", fontsize = 20)
ax[1].set_ylabel("Hubble Residual", fontsize = 20)
ax[0].tick_params(axis='both', labelsize=18)
ax[1].tick_params(axis='both', labelsize=18)
return fig
fig = plot_hubble_distances_and_residuals(zltn_result.point_estimates['mu'], zltn_result.variances['mu'], zcmbs)
plt.show()
fig.savefig("figures/hubble_diagram.pdf", bbox_inches='tight')
fig = plot_hubble_distances_and_residuals(multinormal_result.point_estimates['mu'], multinormal_result.variances['mu'], zcmbs)
plt.show()
fig.savefig("figures/hubble_diagram_multinormal.pdf", bbox_inches='tight')
fig = plot_hubble_distances_and_residuals(laplace_result.point_estimates['mu'], laplace_result.variances['mu'], zcmbs)
plt.show()
fig.savefig("figures/hubble_diagram_laplace.pdf", bbox_inches='tight')
fig = plot_hubble_distances_and_residuals(mcmc_result.point_estimates['mu'], mcmc_result.variances['mu'], zcmbs)
plt.show()
fig.savefig("figures/hubble_diagram_mcmc.pdf", bbox_inches='tight')
print(mcmc_result.stds['mu'])