-
Notifications
You must be signed in to change notification settings - Fork 1
/
StandardModels.py
371 lines (322 loc) · 18.1 KB
/
StandardModels.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
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
# -*- coding: utf-8 -*-
"""
Created on Wed May 31 19:23:33 2017
http://zacstewart.com/2015/04/28/document-classification-with-scikit-learn.html
"""
#from mlxtend.classifier import StackingClassifier
from time import time
from sklearn.svm import SVR,SVC,NuSVR
from sklearn.kernel_ridge import KernelRidge
import xgboost
from sklearn.model_selection import KFold
import numpy as np
from sklearn.linear_model import BayesianRidge,ElasticNetCV,MultiTaskElasticNetCV,RidgeCV
from sklearn.ensemble import GradientBoostingRegressor, GradientBoostingClassifier,RandomForestRegressor
from sklearn.neural_network import MLPRegressor
from sklearn.base import BaseEstimator
from sklearn.metrics.scorer import neg_mean_squared_error_scorer
from sklearn.model_selection import GridSearchCV,RandomizedSearchCV
import warnings
import time
def use_keras_CPU(num_cores=1):
import tensorflow as tf
from keras import backend as K
num_CPU = 1
num_GPU = 0
config = tf.ConfigProto(intra_op_parallelism_threads=num_cores,\
inter_op_parallelism_threads=num_cores, allow_soft_placement=True,\
device_count = {'CPU' : num_CPU, 'GPU' : num_GPU})
session = tf.Session(config=config)
K.set_session(session)
def get_set_count(params):
N=1
for key in params:
N=N*len(params[key])
return N
class MLPRegressorCV(BaseEstimator):
def __init__(self,n_alpha=25,hidden_layer_sizes=None):
self.n_alpha = n_alpha
self.hidden_layer_sizes = hidden_layer_sizes
def fit(self, X, y):
kf = KFold(n_splits=5,shuffle=True, random_state=1)
kff = list(kf.split(X))
errors = 9999*np.ones(self.n_alpha)
alphas = np.logspace(np.log10(1),np.log10(400),self.n_alpha)
for k,alpha in enumerate(alphas):
model = MLPRegressor(activation="relu", solver ="lbfgs",learning_rate ="constant",
learning_rate_init = 0.001, max_iter = 400,random_state = None,
tol = 0.0001, epsilon = 1e-08,alpha=alpha,hidden_layer_sizes=self.hidden_layer_sizes)
err = 0
for train_indices, test_indices in kff:
model.fit(X[train_indices], y[train_indices])
y_pred = model.predict(X[test_indices])
err+=np.mean((y_pred-y[test_indices])**2)
errors[k]=err
ind_best = np.argmax(-errors)
alpha = alphas[ind_best]
print('--> MLPRegressorCV best alpha: %f' % alpha)
self.model = MLPRegressor(activation="relu", solver ="lbfgs",learning_rate ="constant",
learning_rate_init = 0.001, max_iter = 600,random_state = None,
tol = 0.0001, epsilon = 1e-08,alpha=alpha,hidden_layer_sizes=self.hidden_layer_sizes)
self.model.fit(X,y)
return self
def predict(self, X):
return self.model.predict(X)
class KerasENet(BaseEstimator):
def __init__(self, l1_ratio=0.5,alpha=0.01,epochs=1000):
self.l1_ratio = l1_ratio
self.alpha=alpha
self.model = None
self.epochs = epochs
assert 0<=self.l1_ratio<=1,'l1 mixing ratio!'
assert self.alpha>=0,'Bad regularization constant!'
assert self.epochs>5,'Bad epoch value!'
def fit(self, X, y):
from keras.regularizers import l1_l2
from keras.models import Sequential
from keras.layers import Dense
from keras.initializers import Constant
from keras.callbacks import EarlyStopping
from keras import backend as K
K.clear_session()
stop_here = EarlyStopping(patience=5,monitor='loss',min_delta=0.001)
reg = l1_l2(l1=self.alpha*self.l1_ratio,l2=(1.0 - self.l1_ratio)*self.alpha)
init = Constant(value=np.mean(y))
self.model = Sequential()
self.model.add(Dense(1,input_shape=(X.shape[1],),activation='linear',kernel_regularizer=reg,kernel_initializer='glorot_normal',bias_initializer=init))
self.model.compile(loss='mse',optimizer='sgd')
#self.model.fit(X,y,epochs=50,batch_size=X.shape[0],verbose=1) # do at least 50 iterations
self.model.fit(X, y, epochs=20, batch_size=X.shape[0], verbose=0)
self.model.fit(X, y, epochs=self.epochs, batch_size=X.shape[0], verbose=0, callbacks=[stop_here])
return self
def predict(self, X):
return self.model.predict(X)
def get_params(self, deep=True):
# suppose this estimator has parameters "alpha" and "recursive"
return {"alpha": self.alpha, "l1_ratio": self.l1_ratio}
def set_params(self, **parameters):
for parameter, value in parameters.items():
setattr(self,parameter, value)
return self
#def score(self, X, y=None):
# # counts number of values bigger than mean
# return mean_squared_error(self.predict(X),y)
def main(X,Y,Params,print_info=False,is_regression=True,Y_other=None):
parameters = Params['Algorithm'][1]
is_cv_run = False
starttime = time.time()
if print_info:
print('Fitting model \'%s\' for %s' % (Params['Algorithm'][0],'regression' if is_regression else 'classification'))
if Params['Algorithm'][0] =='BayesianRidge':
if not is_regression:
model = BayesianRidge(n_iter=300, tol=0.001,compute_score=False, fit_intercept=True, normalize=False, copy_X=True, verbose=False,**parameters)
#parameters = {'alpha_1': [1e-6,1e-5,1e-4],'alpha_2': [1e-6,1e-5,1e-4], 'lambda_1': [1e-6,1e-5,1e-4], 'lambda_2': [1e-6,1e-5,1e-4]}
else:
model = BayesianRidge(n_iter=300, tol=0.001, compute_score=False, fit_intercept=True, normalize=False, copy_X=True, verbose=False, **parameters)
elif Params['Algorithm'][0] == 'StringKernel':
if not is_regression:
raise (Exception('not implemented'))
else:
# we create an instance of SVM and fit out data.
#
# model = KernelRidge(alpha=parameters['alpha'], kernel='precomputed')
model = SVR(kernel='precomputed', gamma='auto', coef0=0.0, shrinking=True, tol=0.001, cache_size=400, verbose=False, max_iter=-1)
param_grid = {'C': np.logspace(np.log10(0.0001),np.log10(500),25)}
model = NuSVR(kernel='precomputed')#cache_size=400, coef0=0.0, gamma='auto', max_iter=-1, shrinking=True, tol=0.001, verbose=False,**parameters)
param_grid = {'nu':(0.50,)}
model = GridSearchCV(model, param_grid,n_jobs=1, iid=True, refit=True, cv=7, verbose=0,scoring=neg_mean_squared_error_scorer)
is_cv_run = True
elif Params['Algorithm'][0] == 'XGBoost':
# max_depth = 3, learning_rate = 0.1, n_estimators = 100, silent = True, objective = 'reg:linear',
# booster = 'gbtree', n_jobs = 1, nthread = None, gamma = 0, min_child_weight = 1,
# max_delta_step = 0, subsample = 1, colsample_bytree = 1, colsample_bylevel = 1, reg_alpha = 0,
# reg_lambda = 1, scale_pos_weight = 1, base_score = 0.5, random_state = 0, seed = None,
# missing = None
if not is_regression:
model = xgboost.XGBClassifier(missing=None, silent=True,
learning_rate=0.10,
objective='rank:pairwise',
booster='gbtree',
n_jobs=1,
max_delta_step=0,
colsample_bylevel=1,
scale_pos_weight=1,
base_score=0.5,
random_state=666,
colsample_bytree=0.75, # default 1
subsample=0.75,
gamma=0,
reg_alpha=0.01, # default 0
min_child_weight=6,
**parameters)
else:
# model=xgboost.XGBRegressor(missing=None, silent=True,
# learning_rate=0.10,
# objective='reg:linear',#'rank:pairwise' booster='gbtree'
# n_jobs=1,
# booster='gbtree',
# max_delta_step=0,
# colsample_bylevel=1,
# scale_pos_weight=1,
# base_score=0.5,
# random_state=666,
# colsample_bytree=0.75, # default 1
# subsample=0.75,
# gamma=0,
# reg_alpha=0.01, # default 0
# reg_lambda=1.0,
# min_child_weight=6,
# **parameters)
model=xgboost.XGBRegressor(missing=None, silent=True,
learning_rate=0.10,
objective='reg:linear',#'rank:pairwise' booster='gbtree'
n_jobs=1,
booster='gbtree',
random_state=666,
**parameters)
param_grid = {'colsample_bytree': (0.75,1.0),'subsample':(0.75,1.0),'min_child_weight':(3,6,9),'reg_lambda':(0.80,1.0,1.20),'reg_alpha':(0.001,0.01)}
model = GridSearchCV(model, param_grid,n_jobs=1, iid=True, refit=True, cv=7, verbose=0,scoring=neg_mean_squared_error_scorer)
is_cv_run = True
elif Params['Algorithm'][0]== "Keras_ElasticNet":
#use_keras_CPU()
if not is_regression:
raise (Exception('ElasticNet is only for regression!'))
else:
param_grid = {'l1_ratio':(Params['Algorithm'][1]['l1_ratio'],),'alpha':np.logspace(-3,1,15)}
model = GridSearchCV(KerasENet(), param_grid,n_jobs=1, iid=True, refit=True, cv=5, verbose=0,scoring=neg_mean_squared_error_scorer)
# first_output = Dense(1,activation='sigmoid')(first_output)
is_cv_run = True
elif Params['Algorithm'][0] == "Ridge":
if not is_regression:
raise (Exception('Ridge is only for regression!'))
else:
model = RidgeCV(alphas=np.logspace(-1, np.log10(700),parameters['n_alphas']),fit_intercept=True, normalize=False, scoring=None, cv=8, gcv_mode=None, store_cv_values=False)
elif Params['Algorithm'][0]== "ElasticNet":
tol = 0.0001
selection='cyclic'
n_alphas=90
max_iter=1300
if X.shape[1]>4000:
tol = 0.001
selection='random'
n_alphas=60
max_iter=1000
if not is_regression:
raise (Exception('ElasticNet is only for regression!'))
else:
if Params['is_multitarget']:
model = MultiTaskElasticNetCV(eps=0.001, alphas=None, fit_intercept=True, normalize=False, max_iter=max_iter, tol=tol, cv=7, copy_X=True, verbose=0, n_alphas=n_alphas, n_jobs=1, random_state=666, selection=selection, **parameters)
else:
model = ElasticNetCV(eps=0.001,alphas=None,fit_intercept=True,normalize=False,max_iter=max_iter, tol=tol,cv=7, copy_X=True, verbose=0,n_alphas=n_alphas,n_jobs=1,random_state=666,selection=selection,**parameters)
elif Params['Algorithm'][0]== "RandomForest":
if not is_regression:
raise(Exception('not set up (lazy)'))
else:
model = RandomForestRegressor(criterion='mse',min_samples_leaf=1,min_weight_fraction_leaf = 0.0, max_leaf_nodes = None, min_impurity_decrease = 0.0, min_impurity_split = None, bootstrap = True, oob_score = False, n_jobs = 1, random_state = None, verbose = 0, warm_start = False,**parameters)
param_grid = {'max_features': ('auto', 'sqrt'),'min_samples_split':(2,4,),}
model = GridSearchCV(model, param_grid, n_jobs=1, iid=True, refit=True, cv=7, verbose=0, scoring=neg_mean_squared_error_scorer)
is_cv_run = True
elif Params['Algorithm'][0] == 'SVM':
# 0.001, 0.005, 0.01, 0.05, 0.1, 0.5,1.0,1.5,2.0,3.0,4.0,5.0,10.0
if not is_regression:
model = SVC(cache_size=400, coef0=0.0, gamma='auto', max_iter=-1, shrinking=True, tol=0.001, verbose=False,**parameters)
#parameters = {'reg__C':[0.5],'reg__epsilon':[0.1]}
else:
model = SVR(cache_size=400, coef0=0.0, gamma='auto', max_iter=-1, shrinking=True, tol=0.001, verbose=False,**parameters)
param_grid = {'C': np.logspace(np.log10(0.0005),np.log10(10),30)}
#param_grid = {'nu':(0.1,0.3,0.5,0.7,0.9)}
model = GridSearchCV(model, param_grid, n_jobs=1, iid=True, refit=True, cv=8, verbose=0, scoring=neg_mean_squared_error_scorer)
is_cv_run = True
elif Params['Algorithm'][0] == 'GradientBoosting':
if not is_regression:
model = GradientBoostingClassifier(random_state=1,**parameters)
#parameters = {'reg__n_estimators': [140], 'reg__max_depth': [6],'learning_rate':[0.01,0.03,0.1],'min_samples_leaf':[2,3,4]}
else:
model = GradientBoostingRegressor(random_state=1,**parameters)
#parameters = {'reg__n_estimators': [140], 'reg__max_depth': [6]}
elif Params['Algorithm'][0] == 'MLP':
#parameters['hidden_layer_sizes']=[parameters['hidden_layer_sizes']]
#model = MLPRegressorCV(hidden_layer_sizes=parameters['hidden_layer_sizes'])
model = MLPRegressor(activation="relu", solver ="lbfgs",learning_rate ="constant",
learning_rate_init = 0.0011, max_iter = 450,random_state = None,
tol = 0.00013, epsilon = 1e-08,hidden_layer_sizes=parameters['hidden_layer_sizes'])
param_grid = {'alpha': np.logspace(0,np.log10(350),20)}
model = GridSearchCV(model, param_grid, n_jobs=1, iid=True, refit=True, cv=7, verbose=0, scoring=neg_mean_squared_error_scorer)
is_cv_run = True
#model = MLPRegressor(activation="relu", solver ="lbfgs",learning_rate ="constant",
# learning_rate_init = 0.001, power_t = 0.5, max_iter = 500, shuffle = True, random_state = None,
# tol = 0.0001, verbose = False, warm_start = False, momentum = 0.9, epsilon = 1e-08,**parameters)
elif Params['Algorithm'][0] == 'MLP_KERAS':
from keras.models import Sequential
from keras import regularizers
from keras.layers import Dense, Dropout
from keras.callbacks import EarlyStopping
from sklearn.preprocessing import LabelEncoder
from keras.utils import np_utils
import tensorflow as tf
config = tf.ConfigProto()
config.gpu_options.allow_growth = True
session = tf.Session(config=config)
early_stopping = EarlyStopping(monitor='val_loss', patience=5)
model = Sequential()
model.add(Dense(parameters['layers_and_nodes'][0], activation='tanh', input_shape=(X.shape[1],), kernel_initializer='glorot_uniform', kernel_regularizer=regularizers.l2(parameters['l2_regularization']), ))
model.add(Dropout(parameters['dropout'], noise_shape=None, seed=1))
for layer in range(1, len(parameters['layers_and_nodes'])):
model.add(Dense(parameters['layers_and_nodes'][layer], activation='relu', input_shape=(parameters['layers_and_nodes'][layer - 1],),
kernel_initializer='glorot_normal', kernel_regularizer=regularizers.l2(parameters['l2_regularization'])))
model.add(Dropout(parameters['dropout'], noise_shape=None, seed=1))
if not is_regression:
model.add(Dense(1, activation='softmax', input_shape=(parameters['nodes'][-1],)))
model.compile(loss='categorical_crossentropy', optimizer='rmsprop', metrics=['f1'])
encoder = LabelEncoder()
encoder.fit(Y)
encoded_Y = encoder.transform(Y)
# convert integers to dummy variables (i.e. one hot encoded)
Y = np_utils.to_categorical(encoded_Y)
else:
model.add(Dense(1, activation='linear',input_shape=(parameters['layers_and_nodes'][-1],)))
model.compile(loss='mean_squared_error', optimizer='adam', metrics=['mse'])
model.fit(X,Y, batch_size=X.shape[0],epochs=100,validation_split=0,verbose=0)#,callbacks=[early_stopping])
return model
else:
raise(Exception('unknown model'))
#decomposer = LatentDirichletAllocation(n_topics=10, max_iter=10,learning_method='online',learning_offset=50.,random_state=1)
#decomposer = TruncatedSVD(n_components=100,random_state=666)
"""
X = data.iloc[:]['text'].values
y = data.iloc[:]['mylabel'].values.astype(str)
dat = vect.fit_transform(X)
dat = tfidf.fit_transform(dat)
dat = decomposer.fit_transform(dat)
for a in numpy.unique(y):
plt.scatter(dat[y==a,0],dat[y==a,1])
"""
"""
START LOOP
"""
#t0 = time()
# if get_set_count(parameters)>1:
# grid_search = GridSearchCV(model, parameters, n_jobs=6,verbose=1,cv=10,refit=True)
# grid_search.fit(X=X,y=Y)
# best_parameters = grid_search.best_estimator_.get_params()
# print('--> best parameters: %s' % best_parameters)
# return grid_search
# else:
if 1:
start_time = time.time()
print('... training model (X.shape=%s)' % str(X.shape),end='')
warnings.filterwarnings("ignore")
if Y_other is not None and Params['is_multitarget']:
Y = np.expand_dims(Y, axis=1)
model.fit(X=X, y=np.concatenate((Y,Y_other),axis=1))
else:
Y = Y.flatten()
model.fit(X=X,y=Y)
if is_cv_run:
print(' [best gridsearch params: %s] ' % model.best_params_,end='')
if 1:
end_time = time.time()
print(' ... done (%1.1f min)' % ((end_time - start_time)/60.0))
#elapsedtime = (time.time() - starttime) / 60.0
#print('fit done (took %f minutes)' % elapsedtime)
return model