-
Notifications
You must be signed in to change notification settings - Fork 1
/
Predict_Response_Values.py
336 lines (309 loc) · 22.9 KB
/
Predict_Response_Values.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
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
'''
Example usage:
python3 Predict_Response_Values.py Bayesian_Neural_Network 5000 Feature_Matrix_Of_Docking_Scores_And_Number_Of_Occurrences_Of_Substructures.csv Docking_Score
This command works on Linux but not on Windows.
Commands with other models work on Windows.
Tom Lever on 11/15/2023 was able to run
python3 Predict_Response_Values.py Bayesian_Model_Using_BART_Model 265154 Feature_Matrix_Of_Docking_Scores_And_Number_Of_Occurrences_Of_Substructures.csv Docking_Score
on a Rivanna High-Performance computing node with 22 cores, 6 GB RAM per core possibly, or 128 GB host RAM possibly
(Tom entered both the number of cores and the number of gigabytes of memory when setting up an interactive job).
Tom Lever on 11/15/2023 was able to run
python3 Predict_Response_Values.py Bayesian_Model_Using_BART_Model 530307 Feature_Matrix_Of_Docking_Scores_And_Number_Of_Occurrences_Of_Substructures.csv Docking_Score
on a Rivanna High-Performance computing node with 40 cores (the maximum), 6 GB RAM per core possibly, or 256 GB host RAM possibly.
'''
import argparse
import arviz
from ISLP.bart.bart import BART
import cloudpickle
from matplotlib import pyplot as plt
import numpy as np
import pandas as pd
import pymc
import pymc_bart
import pymc.sampling.jax as pmjax
import scipy
from sklearn.linear_model import LinearRegression
from sklearn.preprocessing import StandardScaler
def calculate_one_dimensional_array_of_residual_standard_deviations_slash_errors(
array_of_averages_of_predicted_response_values,
one_dimensional_array_of_response_values_for_testing,
two_dimensional_array_of_values_of_predictors_for_testing
):
array_of_residuals = array_of_averages_of_predicted_response_values - one_dimensional_array_of_response_values_for_testing
number_of_observations = array_of_residuals.shape[0]
number_of_predictors = two_dimensional_array_of_values_of_predictors_for_testing.shape[1]
one_dimensional_array_of_standard_deviations_of_predicted_response_values = np.repeat(
np.sqrt(np.sum(np.square(array_of_residuals)) / (number_of_observations - number_of_predictors - 1)),
number_of_observations
)
return one_dimensional_array_of_standard_deviations_of_predicted_response_values
def main(
model,
number_of_training_or_testing_observations,
path_to_dataset,
response,
should_plot_marginal_posterior_distributions_for_training_data,
should_conduct_posterior_predictive_check
):
random_seed = 0
np.random.seed(random_seed)
feature_matrix = np.loadtxt(path_to_dataset, delimiter = ',', dtype = np.float32, max_rows = 2*number_of_training_or_testing_observations)
column_of_indices = np.arange(0, feature_matrix.shape[0])
print('Feature matrix has shape ' + str(feature_matrix.shape))
print(feature_matrix[0:3, 0:3])
two_dimensional_array_of_values_of_predictors_for_training = feature_matrix[0:number_of_training_or_testing_observations, 1:]
one_dimensional_array_of_response_values_for_training = feature_matrix[0:number_of_training_or_testing_observations, 0]
column_of_indices_for_training = column_of_indices[0:number_of_training_or_testing_observations]
two_dimensional_array_of_values_of_predictors_for_testing = feature_matrix[number_of_training_or_testing_observations:2*number_of_training_or_testing_observations, 1:]
one_dimensional_array_of_response_values_for_testing = feature_matrix[number_of_training_or_testing_observations:2*number_of_training_or_testing_observations, 0]
column_of_indices_for_testing = column_of_indices[number_of_training_or_testing_observations:2*number_of_training_or_testing_observations]
print('Two dimensional array of values of predictors for training has shape ' + str(two_dimensional_array_of_values_of_predictors_for_training.shape))
print(two_dimensional_array_of_values_of_predictors_for_training[0:3, 0:3])
print('One dimensional array of response values for training has shape ' + str(one_dimensional_array_of_response_values_for_training.shape))
print(one_dimensional_array_of_response_values_for_training[0:3])
print('Two dimensional array of values of predictors for testing has shape ' + str(two_dimensional_array_of_values_of_predictors_for_testing.shape))
print(two_dimensional_array_of_values_of_predictors_for_testing[0:3, 0:3])
print('One dimensional array of values of predictors for testing has shape ' + str(one_dimensional_array_of_response_values_for_testing.shape))
print(one_dimensional_array_of_response_values_for_testing[0:3])
for i in range(0, two_dimensional_array_of_values_of_predictors_for_training.shape[1]):
if i % 10 == 0:
print(f'Standardizing column {i}')
if number_of_training_or_testing_observations > 10_000:
random_sample = np.random.choice(two_dimensional_array_of_values_of_predictors_for_training[:, i], 10_000, replace = False)
else:
random_sample = two_dimensional_array_of_values_of_predictors_for_training[:, i]
two_dimensional_array_of_values_of_predictors_for_training[:, i] = (two_dimensional_array_of_values_of_predictors_for_training[:, i] - np.mean(random_sample)) / np.std(random_sample)
if number_of_training_or_testing_observations > 10_000:
random_sample = np.random.choice(two_dimensional_array_of_values_of_predictors_for_testing[:, i], 10_000, replace = False)
else:
random_sample = two_dimensional_array_of_values_of_predictors_for_testing[:, i]
two_dimensional_array_of_values_of_predictors_for_testing[:, i] = (two_dimensional_array_of_values_of_predictors_for_testing[:, i] - np.mean(random_sample)) / np.std(random_sample)
if number_of_training_or_testing_observations > 10_000:
random_sample = np.random.choice(one_dimensional_array_of_response_values_for_training, 10_000, replace = False)
else:
random_sample = one_dimensional_array_of_response_values_for_training
one_dimensional_array_of_response_values_for_training = (one_dimensional_array_of_response_values_for_training - np.mean(random_sample)) / np.std(random_sample)
if number_of_training_or_testing_observations > 10_000:
random_sample = np.random.choice(one_dimensional_array_of_response_values_for_testing, 10_000, replace = False)
else:
random_sample = one_dimensional_array_of_response_values_for_testing
one_dimensional_array_of_response_values_for_testing = (one_dimensional_array_of_response_values_for_testing - np.mean(random_sample)) / np.std(random_sample)
print('Two dimensional array of values of predictors for training has shape ' + str(two_dimensional_array_of_values_of_predictors_for_training.shape))
print(two_dimensional_array_of_values_of_predictors_for_training[0:3, 0:3])
print('One dimensional array of response values for training has shape ' + str(one_dimensional_array_of_response_values_for_training.shape))
print(one_dimensional_array_of_response_values_for_training[0:3])
print('Two dimensional array of values of predictors for testing has shape ' + str(two_dimensional_array_of_values_of_predictors_for_testing.shape))
print(two_dimensional_array_of_values_of_predictors_for_testing[0:3, 0:3])
print('One dimensional array of values of predictors for testing has shape ' + str(one_dimensional_array_of_response_values_for_testing.shape))
print(one_dimensional_array_of_response_values_for_testing[0:3])
if (model == 'BART_Model'):
BART_model = BART(random_state = random_seed)
BART_model.fit(two_dimensional_array_of_values_of_predictors_for_training, one_dimensional_array_of_response_values_for_training)
one_dimensional_array_of_averages_of_predicted_response_values = BART_model.predict(two_dimensional_array_of_values_of_predictors_for_testing)
one_dimensional_array_of_standard_deviations_of_predicted_response_values = calculate_one_dimensional_array_of_residual_standard_deviations_slash_errors(
one_dimensional_array_of_averages_of_predicted_response_values,
one_dimensional_array_of_response_values_for_testing,
two_dimensional_array_of_values_of_predictors_for_testing
)
elif (model.startswith('Bayesian')):
if model == 'Bayesian_Linear_Regression_Model':
with pymc.Model() as pymc_model:
MutableData_of_values_of_predictors = pymc.MutableData('MutableData_of_values_of_predictors', two_dimensional_array_of_values_of_predictors_for_training)
tensor_variable_representing_prior_probability_density_distribution_for_constant_term = pymc.Normal('P(constant term)', mu = 0, sigma = 10)
number_of_predictors = two_dimensional_array_of_values_of_predictors_for_training.shape[1]
tensor_variable_representing_prior_probability_density_distribution_for_vector_of_coefficients = pymc.Normal('P(vector_of_coefficients)', mu = 0, sigma = 10, shape = number_of_predictors)
tensor_variable_representing_expected_value_mu_of_response_values = (
tensor_variable_representing_prior_probability_density_distribution_for_constant_term
+ pymc.math.dot(MutableData_of_values_of_predictors, tensor_variable_representing_prior_probability_density_distribution_for_vector_of_coefficients)
)
tensor_variable_representing_prior_probability_density_distribution_for_standard_deviation = pymc.HalfNormal('P(standard deviation)', sigma = 1)
tensor_variable_representing_likelihood_and_sampling_probability_density_distribution_of_response_values = pymc.Normal(
'P(response value | mu, sigma)',
mu = tensor_variable_representing_expected_value_mu_of_response_values,
sigma = tensor_variable_representing_prior_probability_density_distribution_for_standard_deviation,
observed = one_dimensional_array_of_response_values_for_training
)
inference_data = pmjax.sample_numpyro_nuts(random_seed = random_seed, chain_method = 'vectorized')
elif model == 'Bayesian_Linear_Regression_Model_For_Toy_Dataset':
with pymc.Model() as pymc_model:
MutableData_of_values_of_predictors = pymc.MutableData('MutableData_of_values_of_predictors', two_dimensional_array_of_values_of_predictors_for_training)
tensor_variable_representing_prior_probability_density_distribution_for_constant_term = pymc.Normal('P(constant term)', mu = 1, sigma = 1)
number_of_predictors = two_dimensional_array_of_values_of_predictors_for_training.shape[1]
tensor_variable_representing_prior_probability_density_distribution_for_vector_of_coefficients = pymc.Normal('P(vector_of_coefficients)', mu = [1, 2.5], sigma = 1, shape = number_of_predictors)
tensor_variable_representing_expected_value_mu_of_response_values = (
tensor_variable_representing_prior_probability_density_distribution_for_constant_term
+ pymc.math.dot(MutableData_of_values_of_predictors, tensor_variable_representing_prior_probability_density_distribution_for_vector_of_coefficients)
)
tensor_variable_representing_prior_probability_density_distribution_for_standard_deviation = pymc.HalfNormal('P(standard deviation)', sigma = 1)
tensor_variable_representing_likelihood_and_sampling_probability_density_distribution_of_response_values = pymc.Normal(
'P(response value | mu, sigma)',
mu = tensor_variable_representing_expected_value_mu_of_response_values,
sigma = tensor_variable_representing_prior_probability_density_distribution_for_standard_deviation,
observed = one_dimensional_array_of_response_values_for_training
)
inference_data = pmjax.sample_numpyro_nuts(random_seed = random_seed, chain_method = 'vectorized')
elif model == 'Bayesian_Model_Using_BART_Model':
with pymc.Model() as pymc_model:
MutableData_of_values_of_predictors = pymc.MutableData('MutableData_of_values_of_predictors', two_dimensional_array_of_values_of_predictors_for_training)
tensor_variable_representing_expected_value_mu_of_response_values = pymc_bart.BART(
name = 'mu',
X = MutableData_of_values_of_predictors,
Y = one_dimensional_array_of_response_values_for_training,
m = 50
)
tensor_variable_representing_prior_probability_density_distribution_for_standard_deviation = pymc.HalfNormal('P(sigma)', sigma = 100)
tensor_variable_representing_likelihood_and_sampling_probability_density_distribution_of_response_values = pymc.Normal(
'P(response value | mu, sigma)',
mu = tensor_variable_representing_expected_value_mu_of_response_values,
sigma = tensor_variable_representing_prior_probability_density_distribution_for_standard_deviation,
observed = one_dimensional_array_of_response_values_for_training
)
inference_data = pymc.sample(random_seed = random_seed)
elif model == 'Bayesian_Neural_Network':
# Define the neural network architecture with two hidden layers
input_size = two_dimensional_array_of_values_of_predictors_for_training.shape[1]
hidden_size_1 = 5
hidden_size_2 = 3 # Size of the second hidden layer
# Define the prior for the weights and biases
init_w_1 = np.random.randn(input_size, hidden_size_1)
init_w_2 = np.random.randn(hidden_size_1, hidden_size_2)
init_b_1 = np.zeros(hidden_size_1)
init_b_2 = np.zeros(hidden_size_2)
init_b_3 = np.array([0.0] * hidden_size_2)
with pymc.Model() as pymc_model:
MutableData_of_values_of_predictors = pymc.MutableData('MutableData_of_values_of_predictors', two_dimensional_array_of_values_of_predictors_for_training)
# Weights and biases from input to the first hidden layer
weights_in_1 = pymc.Normal('w_in_1', 0, sigma=1, shape=(input_size, hidden_size_1), initval=init_w_1)
biases_1 = pymc.Normal('b_1', 0, sigma=1, shape=hidden_size_1, initval=init_b_1)
# Weights and biases from the first hidden layer to the second hidden layer
weights_1_2 = pymc.Normal('w_1_2', 0, sigma=1, shape=(hidden_size_1, hidden_size_2), initval=init_w_2)
biases_2 = pymc.Normal('b_2', 0, sigma=1, shape=hidden_size_2, initval=init_b_2)
# Weights and biases from the second hidden layer to the output
weights_2_out = pymc.Normal('w_out_2', 0, sigma=1, shape=hidden_size_2, initval=init_b_3)
# Build the neural network
act_1 = pymc.math.tanh(pymc.math.dot(MutableData_of_values_of_predictors, weights_in_1) + biases_1)
act_2 = pymc.math.tanh(pymc.math.dot(act_1, weights_1_2) + biases_2)
tensor_variable_representing_expected_value_mu_of_response_values = pymc.math.dot(act_2, weights_2_out)
range_of_observed_response_values = max(one_dimensional_array_of_response_values_for_training) - min(one_dimensional_array_of_response_values_for_training)
tensor_variable_representing_expected_value_mu_of_response_values = tensor_variable_representing_expected_value_mu_of_response_values * range_of_observed_response_values
tensor_variable_representing_prior_probability_density_distribution_for_standard_deviation = pymc.HalfNormal('P(sigma)', sigma = 100)
# Likelihood (sampling distribution) of the target values
tensor_variable_representing_likelihood_and_sampling_probability_density_distribution_of_response_values = pymc.Normal(
'P(response value | mu, sigma)',
mu = tensor_variable_representing_expected_value_mu_of_response_values,
sigma = tensor_variable_representing_prior_probability_density_distribution_for_standard_deviation,
observed = one_dimensional_array_of_response_values_for_training
)
inference_data = pmjax.sample_numpyro_nuts(random_seed = random_seed, chain_method = 'vectorized')
'''
with open(f'Pickled_{model}.mod', 'wb') as file:
cloudpickle.dump(pymc_model, file)
with open(f'Pickled_Training_Inference_Data_For_{model}.dat', 'wb') as file:
cloudpickle.dump(inference_data, file)
'''
if should_plot_marginal_posterior_distributions_for_training_data:
arviz.plot_trace(inference_data)
plt.savefig('Marginal_Posterior_Distributions_For_Training_Data.png')
with pymc_model:
inference_data_for_posterior_predictive_probability_density_distribution_for_training_data = pymc.sample_posterior_predictive(
trace = inference_data,
random_seed = random_seed
)
pymc.set_data({'MutableData_of_values_of_predictors': two_dimensional_array_of_values_of_predictors_for_testing})
inference_data_for_posterior_predictive_probability_density_distribution_for_testing_data = pymc.sample_posterior_predictive(
trace = inference_data,
random_seed = random_seed
)
array_of_predicted_response_values = inference_data_for_posterior_predictive_probability_density_distribution_for_testing_data.posterior_predictive['P(response value | mu, sigma)']
one_dimensional_array_of_averages_of_predicted_response_values = array_of_predicted_response_values.mean(axis = (0, 1))
one_dimensional_array_of_standard_deviations_of_predicted_response_values = array_of_predicted_response_values.std(axis = (0, 1))
if should_conduct_posterior_predictive_check:
fig, ax = plt.subplots(
nrows = 2,
ncols = 1,
figsize = (8, 7),
sharex = True,
sharey = True,
layout = 'constrained'
)
arviz.plot_ppc(
data = inference_data_for_posterior_predictive_probability_density_distribution_for_training_data,
observed_rug = True,
ax = ax[0]
)
ax[0].set(
title = 'Posterior Predictive Check For Training Data',
)
arviz.plot_ppc(
data = inference_data_for_posterior_predictive_probability_density_distribution_for_testing_data,
observed_rug = True,
ax = ax[1]
)
ax[1].set(
title = 'Posterior Predictive Check For Testing Data'
)
plt.legend(loc = 'upper right')
plt.savefig('Posterior_Predictive_Checks_For_Training_And_Testing_Data.png')
elif (model == 'Linear_Regression_Model'):
linear_regression_model = LinearRegression()
linear_regression_model.fit(two_dimensional_array_of_values_of_predictors_for_training, one_dimensional_array_of_response_values_for_training)
one_dimensional_array_of_averages_of_predicted_response_values = linear_regression_model.predict(two_dimensional_array_of_values_of_predictors_for_testing)
one_dimensional_array_of_standard_deviations_of_predicted_response_values = calculate_one_dimensional_array_of_residual_standard_deviations_slash_errors(
one_dimensional_array_of_averages_of_predicted_response_values,
one_dimensional_array_of_response_values_for_testing,
two_dimensional_array_of_values_of_predictors_for_testing
)
else:
raise Exception(f'We cannot predict response values for type of model "{model}"')
data_frame_of_observed_response_values_and_averages_and_standard_deviations_of_predicted_response_values = pd.DataFrame(
{
'index': column_of_indices_for_testing,
'observed_response_value': one_dimensional_array_of_response_values_for_testing,
'average_of_predicted_response_values': one_dimensional_array_of_averages_of_predicted_response_values,
'standard_deviation_of_predicted_response_values': one_dimensional_array_of_standard_deviations_of_predicted_response_values
}
)
data_frame_of_observed_response_values_and_averages_and_standard_deviations_of_predicted_response_values.to_csv(
'Data_Frame_Of_Observed_Response_Values_And_Averages_And_Standard_Deviations_Of_Predicted_Response_Values.csv',
index = False
)
if __name__ == '__main__':
parser = argparse.ArgumentParser(prog = 'Predict Response Values', description = 'This program predicts response values.')
parser.add_argument(
'model',
choices = [
'BART_Model',
'Bayesian_Linear_Regression_Model',
'Bayesian_Linear_Regression_Model_For_Toy_Dataset',
'Bayesian_Model_Using_BART_Model',
'Bayesian_Neural_Network',
'Linear_Regression_Model'
],
help = 'type of model to train'
)
parser.add_argument('number_of_training_or_testing_observations', help = 'number of training or testing observations')
parser.add_argument('path_to_dataset', help = 'path to dataset')
parser.add_argument('response', help = 'response')
parser.add_argument('--should_plot_marginal_posterior_distributions_for_training_data', action = 'store_true', help = 'should plot marginal posterior probability density distributions for training data')
parser.add_argument('--should_conduct_posterior_predictive_check', action = 'store_true', help = 'should conduct posterior predictive checks')
args = parser.parse_args()
model = args.model
number_of_training_or_testing_observations = int(args.number_of_training_or_testing_observations)
path_to_dataset = args.path_to_dataset
response = args.response
should_plot_marginal_posterior_distributions_for_training_data = args.should_plot_marginal_posterior_distributions_for_training_data
should_conduct_posterior_predictive_check = args.should_conduct_posterior_predictive_check
print(f'model: {model}')
print(f'number of training or testing observations: {number_of_training_or_testing_observations}')
print(f'path to dataset: {path_to_dataset}')
print(f'response: {response}')
print(f'should plot marginal posterior distributions for training data: {should_plot_marginal_posterior_distributions_for_training_data}')
print(f'should conduct posterior predictive check: {should_conduct_posterior_predictive_check}')
main(
model = model,
number_of_training_or_testing_observations = number_of_training_or_testing_observations,
path_to_dataset = path_to_dataset,
response = response,
should_plot_marginal_posterior_distributions_for_training_data = should_plot_marginal_posterior_distributions_for_training_data,
should_conduct_posterior_predictive_check = should_conduct_posterior_predictive_check
)