-
Notifications
You must be signed in to change notification settings - Fork 0
/
nn.py
125 lines (99 loc) · 3.87 KB
/
nn.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
""""""
import numpy as np
import tensorflow as tf
import json
import pandas as pd
with open("test.json") as fd:
data = json.load(fd)
simple_headings = ['C1R', 'C1S', 'C2R', 'C2S',
'Stack1', 'Stack2', 'Stack3', 'Stack4', 'Stack5', 'Stack6']
shaped_headings = ['C1R', 'C2R', 'ShapeP', 'ShapeS', 'ShapeO',
'Stack1', 'Stack2', 'Stack3', 'Stack4', 'Stack5', 'Stack6']
one_hot_everything_headings = ['R1' + str(x) for x in range(13)] + ['R2' + str(x) for x in range(13)] + ['ShapeP', 'ShapeS', 'ShapeO'] + ['Stack1', 'Stack2', 'Stack3', 'Stack4', 'Stack5', 'Stack6']
# data = [x + [y] for x,y in data]
data = pd.DataFrame(np.array(data), columns=one_hot_everything_headings + ['wasRaise'])
# data.drop_duplicates(subset = one_hot_everything_headings,
# keep = False, inplace = True)
print(data.head())
print(data.shape)
# input()
mask = np.random.rand(len(data)) < 0.85
train_data = data[mask]
test_data = data[~mask]
def extract_data_labels(data):
labels = data['wasRaise'].values
data_without_string_columns_or_sale_price = data.copy().drop(columns=['wasRaise'])
data = data_without_string_columns_or_sale_price.values.astype('float64')
return (data, labels)
(train_data, train_targets) = extract_data_labels(train_data)
(test_data, test_targets) = extract_data_labels(test_data)
# MODEL
# import ipdb; ipdb.set_trace()
from tensorflow.keras import models
from tensorflow.keras import layers
_, data_cols = train_data.shape
λ = 1
k = 1
num_epochs = 1000
β = 0.9
β_2 = 0.99
def build_model():
# Because we will need to instantiate
# the same model multiple times,
# we use a function to construct it.
# act = tf.keras.activations.relu(x, alpha=0.01, max_value=None, threshold=0.0)
model = models.Sequential()
# model.add(layers.Dense(data_cols * 2, activation='relu',
# input_shape=(train_data.shape[1],)))
# model.add(layers.Dense(data_cols, activation='relu'))
# model.add(layers.Dense(data_cols//2, activation='relu'))
# model.add(layers.Dense(1, activation='relu'))
model.add(layers.Dense(data_cols * 4, activation='relu',
input_shape=(train_data.shape[1],)))
model.add(layers.Dense(data_cols * 4, activation='relu'))
model.add(layers.Dense(data_cols * 2, activation='relu'))
model.add(layers.Dense(data_cols * 2, activation='relu'))
model.add(layers.Dense(data_cols, activation='relu'))
model.add(layers.Dense(data_cols, activation='relu'))
model.add(layers.Dense(1, activation='sigmoid')) # should this be sigmoid?
# bin_ce
filepath = "saved_net.py"
checkpoint = tf.keras.callbacks.ModelCheckpoint(filepath, monitor='val_accuracy', verbose=1, save_best_only=True, mode='max')
callbacks_list = [checkpoint]
opt = tf.keras.optimizers.Adam(learning_rate=λ, beta_1=β, beta_2=β_2)
# model.compile(optimizer=opt, loss='binary_crossentropy', metrics=['accuracy'])
model.compile(optimizer=opt, loss='mae', metrics=['accuracy', 'mae', 'binary_accuracy'])
return model
model = build_model()
history = model.fit(train_data, train_targets,
validation_data=(test_data, test_targets),
epochs=num_epochs, batch_size=10000, verbose=1)
res = model.evaluate(test_data, test_targets, verbose=0)
print(res)
print('keys:', history.history.keys())
for i in range(13):
print(i)
for j in range(13):
for s in [[1,0,0], [0,1,0], [0,0,1]]:
r1 = list(np.zeros((13, )))
r2 = list(np.zeros((13, )))
r1[i] = 1
r2[j] = 1
if i >= j:
print(i,j,s)
print(model.predict([r1 + r2 + s + [25,25,25,25,25,25]]))
def show_rank(r): {
0: '1',
1: '2',
2: '3',
3: '4',
4: '5',
5: '6',
6: '7',
7: '8',
8: '9',
9: 'J',
10: 'Q',
11: 'K',
12: 'A',
}