-
Notifications
You must be signed in to change notification settings - Fork 8
/
plots.py
33 lines (29 loc) · 1.31 KB
/
plots.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
import matplotlib.pyplot as plt
import numpy as np
def plot_bayes(res,phi_range, ax=plt):
smooth_phi = np.linspace(*phi_range,num = 200).reshape(-1,1)
try:
y,std = res.models[-1].predict(smooth_phi, return_std = True)
y,std = -y,-std
ax.plot(smooth_phi[:,0],y, c = 'k',label = 'GP')
ax.plot(smooth_phi[:,0],y+std, c = 'k', linestyle = 'dashed')
ax.plot(smooth_phi[:,0],y-std, c = 'k', linestyle = 'dashed')
ax.fill_between(smooth_phi[:,0],y-std,y+std, color = 'k', alpha = 0.2)
except IndexError:
pass
ax.scatter(res.x_iters,-res.func_vals, marker = '.', s = 100, color = 'k')
ax.axvline(res.x_iters[-1],color = 'b', label = 'last x')
ax.axvline(res.x, c = 'r', label = 'opt. x')
ax.set_xlim(*phi_range)
ax.legend()
def plot_data(data, data_range, ax = plt):
bins = np.linspace(*data_range, num = 11)
ax.hist(data, bins = bins, label = 'data')
ax.legend()
def plot_posterior(prior,posterior, best_theta, true_theta, theta_range, map_bins = 20, ax = plt):
xs = np.linspace(*theta_range,num = map_bins)
pri = ax.plot(xs,prior.pdf(xs), label = 'prior')
pos = ax.plot(xs,posterior.pdf(xs), label = 'posterior')
ax.axvline(best_theta, c = 'k', label = 'MAP')
ax.axvline(true_theta, c = 'grey', label = 'truth')
ax.legend()