/
dropout.py
99 lines (69 loc) · 3.57 KB
/
dropout.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
import lightwood
import pandas as pd
import numpy as np
import random
from collections import Counter
random.seed(66)
def gen_multiply():
n = 1000
# generate random numbers between -10 and 10
data_train = {'x': [random.randint(-15, 5) for i in range(n)],
'y': [random.randint(-15, 5) for i in range(n)]}
data_test = {'x': [random.randint(-5, 15) for i in range(n)],
'y': [random.randint(-5, 15) for i in range(n)]}
# target variable to be the multiplication of the two
data_train['z'] = [data_train['x'][i] * data_train['y'][i] for i in range(n)]
data_test['z'] = [data_test['x'][i] * data_test['y'][i] for i in range(n)]
df_train = pd.DataFrame(data_train)
df_test = pd.DataFrame(data_test)
return (df_train, df_test, [['x'],['y']], 'z','multiplied')
def gen_correlate():
n = 500
train = True
data_train = {}
data_test = {}
for data, nr_ele in [(data_train,n), (data_test,n)]:
for i in range(1,5):
data[f'x_{i}'] = [random.random()*50 + 25 for _ in range(nr_ele)]
data['y'] = [data['x_1'][i] * 0.9 + data['x_2'][i] * 0.09 + data['x_3'][i] * 0.009 + data['x_4'][i] * 0.0009 for i in range(nr_ele)]
data_train = pd.DataFrame(data_train)
data_test = pd.DataFrame(data_test)
return (data_train, data_test, [['x_1'],['x_2','x_4'],['x_3'],['x_2']], 'y','correlated')
def gen_categorical():
n = 1000
nr_inputs = 6
options = ['a','b','c']
data_train = {}
data_test = {}
for data, nr_ele in [(data_train,n), (data_test,int(n/5))]:
for i in range(nr_inputs):
data[f'x_{i}'] = [random.choice(options) for _ in range(nr_ele)]
data['y'] = [Counter([data[f'x_{i}'][n] for i in range(nr_inputs)]).most_common(1)[0][0] for n in range(nr_ele)]
data_train = pd.DataFrame(data_train)
data_test = pd.DataFrame(data_test)
return (data_train, data_test, [['x_1'],['x_2','x_4'],['x_5','x_4','x_3']], 'y','cummulative')
def iter_function(epoch, training_error, test_error, test_error_gradient, test_accuracy):
print(f'Epoch: {epoch}, Train Error: {training_error}, Test Error: {test_error}, Test Error Gradient: {test_error_gradient}, Test Accuracy: {test_accuracy}')
test_cases = [gen_multiply(),gen_correlate(),gen_categorical()]
log_map = {}
for i, data in enumerate(test_cases):
df_train, df_test, dropout_arr, out_col, name = data
pmap = {}
accmap = {}
pmap['normal'] = lightwood.Predictor(output=[out_col])
pmap['normal'].learn(from_data=df_train, callback_on_iter=iter_function, eval_every_x_epochs=100)
accmap['normal'] = pmap['normal'].calculate_accuracy(from_data=df_test)[out_col]['value']
for cols in dropout_arr:
mk = 'missing_' + '_'.join(cols)
pmap[mk] = lightwood.Predictor(output=[out_col])
pmap[mk].learn(from_data=df_train.drop(columns=cols), callback_on_iter=iter_function, eval_every_x_epochs=100)
accmap[mk + '_unfit'] = pmap['normal'].calculate_accuracy(from_data=df_test.drop(columns=cols))[out_col]['value']
accmap[mk + '_fit'] = pmap[mk].calculate_accuracy(from_data=df_test.drop(columns=cols))[out_col]['value']
text = f'\n---------\nTest case {name}\n---------\nNormal accuracy of: ' + str(accmap['normal'])
for cols in dropout_arr:
mk = 'missing_' + '_'.join(cols)
text += f'\nSpecially-trained trained accuracy when {cols} missing: ' + str(accmap[mk + '_fit'])
text += f'\nNormally-trained trained accuracy when {cols} missing: ' + str(accmap[mk + '_unfit'])
log_map[name] = text
for k in log_map:
print(log_map[k])