-
Notifications
You must be signed in to change notification settings - Fork 13
/
2d_classification.py
128 lines (106 loc) · 4.65 KB
/
2d_classification.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
import sys
sys.path.insert(0, '../')
import numpy as np
from jax.experimental import optimizers
import matplotlib.pyplot as plt
from matplotlib.colors import hsv_to_rgb, rgb_to_hsv, ListedColormap
from scipy.interpolate import interp1d
import time
from sde_gp import SDEGP
import approximate_inference as approx_inf
import priors
import likelihoods
from utils import softplus_list, plot_2d_classification, plot_2d_classification_filtering
plot_intermediate = False
print('loading banana data ...')
inputs = np.loadtxt('../../data/banana_X_train', delimiter=',')
X = inputs[:, :1] # temporal inputs (x-axis)
R = inputs[:, 1:] # spatial inputs (y-axis)
Y = np.loadtxt('../../data/banana_Y_train')[:, None] # observations / labels
# Test points
Xtest, Rtest = np.mgrid[-2.8:2.8:100j, -2.8:2.8:100j]
# Xtest = np.vstack((Xtest.flatten(), Ytest.flatten())).T
# X0test, X1test = np.linspace(-3., 3., num=100), np.linspace(-3., 3., num=100)
# plot_2d_classification(None, 0)
np.random.seed(99)
N = X.shape[0] # number of training points
var_f = 0.3 # GP variance
len_time = 0.3 # temporal lengthscale
len_space = 0.3 # spacial lengthscale
prior = priors.SpatioTemporalMatern52(variance=var_f, lengthscale_time=len_time, lengthscale_space=len_space)
lik = likelihoods.Probit()
inf_method = approx_inf.ExpectationPropagation(power=0.5)
# inf_method = approx_inf.StatisticallyLinearisedEP()
# inf_method = approx_inf.ExtendedKalmanSmoother()
# inf_method = approx_inf.VariationalInference()
model = SDEGP(prior=prior, likelihood=lik, t=X, y=Y, r=R, approx_inf=inf_method)
opt_init, opt_update, get_params = optimizers.adam(step_size=2e-1)
# parameters should be a 2-element list [param_prior, param_likelihood]
opt_state = opt_init([model.prior.hyp, model.likelihood.hyp])
def gradient_step(i, state, mod, plot_num_, mu_prev_):
params = get_params(state)
mod.prior.hyp = params[0]
mod.likelihood.hyp = params[1]
# grad(Filter) + Smoother:
neg_log_marg_lik, gradients = mod.run()
# neg_log_marg_lik, gradients = mod.run_two_stage() # <-- less elegant but reduces compile time
prior_params = softplus_list(params[0])
print('iter %2d: var=%1.2f len_time=%1.2f len_space=%1.2f, nlml=%2.2f' %
(i, prior_params[0], prior_params[1], prior_params[2], neg_log_marg_lik))
if plot_intermediate:
plot_2d_classification(mod, i)
# plot_num_, mu_prev_ = plot_2d_classification_filtering(mod, i, plot_num_, mu_prev_)
return opt_update(i, gradients, state), plot_num_, mu_prev_
plot_num = 0
mu_prev = None
print('optimising the hyperparameters ...')
t0 = time.time()
for j in range(50):
opt_state, plot_num, mu_prev = gradient_step(j, opt_state, model, plot_num, mu_prev)
t1 = time.time()
print('optimisation time: %2.2f secs' % (t1-t0))
# calculate posterior predictive distribution via filtering and smoothing at train & test locations:
print('calculating the posterior predictive distribution ...')
t0 = time.time()
mu, var = model.predict(t=Xtest, r=Rtest)
t1 = time.time()
print('prediction time: %2.2f secs' % (t1-t0))
link_fn = model.likelihood.link_fn
print('plotting ...')
plt.figure(1)
for label, mark in [[1, 'o'], [0, 'o']]:
ind = Y[:, 0] == label
# ax.plot(X[ind, 0], X[ind, 1], mark)
plt.scatter(X[ind], R[ind], s=50, alpha=.5)
# ax.imshow(mu.T)
plt.contour(Xtest, Rtest, mu, levels=[.0], colors='k', linewidths=4.)
# plt.axis('equal')
plt.tick_params(axis='x', which='both', bottom=False, top=False, labelbottom=False)
plt.tick_params(axis='y', which='both', right=False, left=False, labelleft=False)
# ax.axis('off')
lim = 2.8
plt.xlim(-lim, lim)
plt.ylim(-lim, lim)
# plt.savefig('output/data.png')
# x1 = np.linspace(-lim, lim, num=100)
# x2 = np.linspace(-lim, lim, num=100)
cmap_ = [[1, 0.498039215686275, 0.0549019607843137], [0.12156862745098, 0.466666666666667, 0.705882352941177]]
cmap = hsv_to_rgb(interp1d([0., 1.], rgb_to_hsv(cmap_), axis=0)(link_fn(np.linspace(-3.5, 3.5, num=64))))
newcmp = ListedColormap(cmap)
plt.figure(2)
im = plt.imshow(link_fn(mu).T, cmap=newcmp, extent=[-lim, lim, -lim, lim], origin='lower')
cb = plt.colorbar(im)
cb.set_ticks([cb.vmin, 0, cb.vmax])
cb.set_ticklabels([-1, 0, 1])
plt.contour(Xtest, Rtest, mu, levels=[.0], colors='k', linewidths=1.5)
# plt.axis('equal')
for label in [1, 0]:
ind = Y[:, 0] == label
plt.scatter(X[ind], R[ind], s=50, alpha=.5, edgecolor='k')
# plt.title('Iteration: %02d' % (j + 1), loc='right', fontweight='bold')
plt.tick_params(axis='x', which='both', bottom=False, top=False, labelbottom=False)
plt.tick_params(axis='y', which='both', right=False, left=False, labelleft=False)
plt.xlim(-lim, lim)
plt.ylim(-lim, lim)
# plt.savefig('output/output_%04d.png' % 1600)
plt.show()