-
Notifications
You must be signed in to change notification settings - Fork 0
/
plot_data.py
154 lines (116 loc) · 5.29 KB
/
plot_data.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
import pandas as pd
import math
import torch
import gpytorch
from matplotlib import pyplot as plt
import numpy as np
import seaborn as sns
from sklearn.metrics.regression import r2_score
# setting up model
class ExactGPModel(gpytorch.models.ExactGP):
def __init__(self, train_x, train_y, likelihood):
super(ExactGPModel, self).__init__(train_x, train_y, likelihood)
self.mean_module = gpytorch.means.ConstantMean()
self.covar_module = gpytorch.kernels.ScaleKernel(gpytorch.kernels.MaternKernel(ard_num_dims=train_x.shape[1]) +
gpytorch.kernels.RBFKernel(ard_num_dims=train_x.shape[1]))
def forward(self, x):
mean_x = self.mean_module(x)
covar_x = self.covar_module(x)
return gpytorch.distributions.MultivariateNormal(mean_x, covar_x)
class GPMixturModel(gpytorch.models.ExactGP):
def __init__(self, train_x, train_y, likelihood):
super(GPMixturModel, self).__init__(train_x, train_y, likelihood)
self.mean_module = gpytorch.means.ConstantMean()
self.covar_module = gpytorch.kernels.ScaleKernel(
gpytorch.kernels.SpectralMixtureKernel(num_mixtures=4, ard_num_dims=train_x.shape[1])+
gpytorch.kernels.RBFKernel(ard_num_dims=train_x.shape[1])
)
def forward(self, x):
mean_x = self.mean_module(x)
covar_x = self.covar_module(x)
return gpytorch.distributions.MultivariateNormal(mean_x, covar_x)
def train(train_x_pre, train_y_pre, training_iter=100):
train_x = train_x_pre.cuda()
train_y = train_y_pre.cuda()
likelihood = gpytorch.likelihoods.GaussianLikelihood()
model = ExactGPModel(train_x, train_y, likelihood)
likelihood = likelihood.cuda()
model = model.cuda()
# training the model
# Find optimal model hyperparameters
model.train()
likelihood.train()
# adamで最適化
optimizer = torch.optim.Adam([
{"params": model.parameters()},
], lr=0.05)
# defining loss og GPs (marginal likelihood)
mll = gpytorch.mlls.ExactMarginalLogLikelihood(likelihood, model)
for i in range(training_iter):
optimizer.zero_grad()
output = model(train_x)
loss = -mll(output, train_y)
loss.backward()
print(f"Iter {i + 1}/{training_iter} - Loss: {loss.item():.3f} noise:{model.likelihood.noise.item():.3f}")
optimizer.step()
return model, likelihood
def train2(train_x_pre, train_y_pre, training_iter=100):
train_x = train_x_pre.cuda()
train_y = train_y_pre.cuda()
likelihood = gpytorch.likelihoods.GaussianLikelihood()
model = GPMixturModel(train_x, train_y, likelihood)
likelihood = likelihood.cuda()
model = model.cuda()
# training the model
# Find optimal model hyperparameters
model.train()
likelihood.train()
# adamで最適化
optimizer = torch.optim.Adam([
{"params": model.parameters()},
], lr=0.05)
# defining loss og GPs (marginal likelihood)
mll = gpytorch.mlls.ExactMarginalLogLikelihood(likelihood, model)
for i in range(training_iter):
optimizer.zero_grad()
output = model(train_x)
loss = -mll(output, train_y)
loss.backward()
print(f"Iter {i + 1}/{training_iter} - Loss: {loss.item():.3f} noise:{model.likelihood.noise.item():.3f}")
optimizer.step()
return model, likelihood
def test(test_x, model, likelihood):
model.eval()
likelihood.eval()
test_x = test_x.cuda()
# Te
# st points are regularly spaced along [0,1]
# Make predictions by feeding model through likelihood
with torch.no_grad(), gpytorch.settings.fast_pred_var():
observed_pred = likelihood(model(test_x))
mean = observed_pred.mean
lower, upper = observed_pred.confidence_region()
return observed_pred, mean, lower, upper
def main():
path1 = "analysis/wine/log_data.csv"
path2 = "analysis/wine/pc_data.csv"
df1 = pd.read_csv(path1)
df2 = pd.read_csv(path2)
np.random.seed(777)
randint = np.random.choice(df1.index, size=250)
test_x = torch.from_numpy(df1.loc[randint].drop('quality', axis=1).values).float()
test_y = torch.from_numpy(df1.loc[randint, 'quality'].values).float()
train_x_pre = torch.from_numpy(df1.drop(randint).drop('quality', axis=1).values).float()
train_y_pre = torch.from_numpy(df1.drop(randint).loc[:,'quality'].values).float()
model, likelihood = train(train_x_pre, train_y_pre, training_iter=1000)
observed_pred, mean, lower, upper = test(test_x, model, likelihood)
sns.scatterplot(x=test_y.numpy(), y=mean.cpu().numpy())
r2_score(test_y.numpy(), mean.cpu().numpy())
# test_x2 = torch.from_numpy(df2.loc[randint].drop('quality', axis=1).values).float()
# test_y2 = torch.from_numpy(df2.loc[randint, 'quality'].values).float()
# train_x_pre2 = torch.from_numpy(df2.drop(randint).drop('quality', axis=1).values).float()
# train_y_pre2 = torch.from_numpy(df2.drop(randint).loc[:,'quality'].values).float()
model2, likelihood2 = train2(train_x_pre, train_y_pre, training_iter=1000)
observed_pred2, mean2, lower2, upper2 = test(test_x, model2, likelihood2)
sns.scatterplot(x=test_y.numpy(), y=mean2.cpu().numpy())
r2_score(test_y.numpy(), mean2.cpu().numpy())