generate_DeepConvLSTM_model() - Code Metrics - Inspection of "Add comment to architecture test" - NLeSC/mcfly - Measure and Improve Code Quality continuously with Scrutinizer

Completed

Push — master ( a63bb3...d33a20 )

by Christiaan

created 2016-07-27 12:45 UTC

generate_DeepConvLSTM_model() B

↳ Parent: Project

Complexity

Conditions

Size

Total Lines

Duplication

Lines	0
Ratio	0 %

Code Coverage

Tests	20
CRAP Score	4

Importance

Changes	6
Bugs	1	Features	3

Metric	Value
cc	4
c	6
b	1
f	3
dl	0
loc	45
ccs	20
cts	20
cp	1
crap	4
rs	8.5806

from keras.models import Sequential
from keras.layers import Dense, Activation, Convolution1D, Flatten, MaxPooling1D, Lambda, Convolution2D, Flatten, Reshape, LSTM, Dropout, TimeDistributed, Permute, BatchNormalization

from keras.regularizers import l2
from keras.optimizers import Adam
import numpy as np


def generate_models(x_shape, number_of_classes, number_of_models=5, model_type=None, **kwargs):

    """ Generate one or multiple Keras models with random (default), or predefined, hyperparameters."""

    models = []
    for _ in range(0, number_of_models):
        if model_type is None:  # random model choice:
            current_model_type = 'CNN' if np.random.random(
            ) < 0.5 else 'DeepConvLSTM'
        else:  # user-defined model choice:
            current_model_type = model_type

        if current_model_type == 'CNN':
            generate_model = generate_CNN_model  # object is a function
            generate_hyperparameter_set = generate_CNN_hyperparameter_set
        if current_model_type == 'DeepConvLSTM':
            generate_model = generate_DeepConvLSTM_model  # object is a function

            generate_hyperparameter_set = generate_DeepConvLSTM_hyperparameter_set

        hyperparameters = generate_hyperparameter_set(**kwargs)
        models.append(
            (generate_model(x_shape, number_of_classes, **hyperparameters), hyperparameters, current_model_type))

    return models


def generate_DeepConvLSTM_model(x_shape, class_number, filters, lstm_dims, learning_rate=0.01,

                                regularization_rate=0.01):
    """
    Generate a model with convolution and LSTM layers.
    See Ordonez et al., 2016, http://dx.doi.org/10.3390/s16010115

    The compiled Keras model is returned.
    """
    dim_length = x_shape[1]  # number of samples in a time series
    dim_channels = x_shape[2]  # number of channels
    output_dim = class_number  # number of classes
    weightinit = 'lecun_uniform'  # weight initialization
    model = Sequential()  # initialize model
    model.add(BatchNormalization(input_shape=(dim_length, dim_channels)))
    # reshape a 2 dimensional array per file/person/object into a
    # 3 dimensional array
    model.add(
        Reshape(target_shape=(1, dim_length, dim_channels)))
    for filt in filters:
        # filt: number of filters used in a layer
        # filters: vector of filt values
        model.add(
            Convolution2D(filt, nb_row=3, nb_col=1, border_mode='same',
                          W_regularizer=l2(regularization_rate), init=weightinit))

        model.add(Activation('relu'))
    # reshape 3 dimensional array back into a 2 dimensional array,
    # but now with more dept as we have the the filters for each channel
    model.add(Reshape(target_shape=(dim_length, filters[-1] * dim_channels)))

    for lstm_dim in lstm_dims:
        model.add(LSTM(output_dim=lstm_dim, return_sequences=True,
                       activation='tanh'))

    model.add(Dropout(0.5))  # dropout before the dense layer
    # set up final dense layer such that every timestamp is given one classification

    model.add(TimeDistributed(Dense(output_dim, W_regularizer=l2(regularization_rate))))

    model.add(Activation("softmax"))
    # Final classification layer - per timestep
    model.add(Lambda(lambda x: x[:, -1, :], output_shape=[output_dim]))

    model.compile(loss='categorical_crossentropy',
                  optimizer=Adam(lr=learning_rate),
                  metrics=['accuracy'])

    return model


def generate_CNN_model(x_shape, class_number, filters, fc_hidden_nodes,

                       learning_rate=0.01, regularization_rate=0.01):
    """
    Generate a convolutional neural network (CNN) model.

    The compiled Keras model is returned.
    """
    dim_length = x_shape[1]  # number of samples in a time series
    dim_channels = x_shape[2]  # number of channels
    outputdim = class_number  # number of classes
    weightinit = 'lecun_uniform'  # weight initialization
    model = Sequential()
    model.add(BatchNormalization(input_shape=(dim_length, dim_channels)))
    for filter_number in filters:
        model.add(Convolution1D(filter_number, 3, border_mode='same',
                                W_regularizer=l2(regularization_rate), init=weightinit))

        model.add(Activation('relu'))
    model.add(Flatten())
    model.add(Dense(output_dim=fc_hidden_nodes,
                    W_regularizer=l2(regularization_rate), init=weightinit))  # Fully connected layer

    model.add(Activation('relu'))  # Relu activation
    model.add(Dense(output_dim=outputdim, init=weightinit))
    model.add(Activation("softmax"))  # Final classification layer

    model.compile(loss='categorical_crossentropy',
                  optimizer=Adam(lr=learning_rate),
                  metrics=['accuracy'])

    return model


def generate_CNN_hyperparameter_set(min_layers=1, max_layers=10,

                                    min_filters=10, max_filters=100,
                                    min_fc_nodes=10, max_fc_nodes=100):
    """ Generate a hyperparameter set that define a CNN model."""
    hyperparameters = generate_base_hyper_parameter_set()
    number_of_layers = np.random.randint(min_layers, max_layers)
    hyperparameters['filters'] = np.random.randint(
        min_filters, max_filters, number_of_layers)
    hyperparameters['fc_hidden_nodes'] = np.random.randint(
        min_fc_nodes, max_fc_nodes)
    return hyperparameters


def generate_DeepConvLSTM_hyperparameter_set(

    min_conv_layers=1, max_conv_layers=10,
    min_conv_filters=10, max_conv_filters=100,
    min_lstm_layers=1, max_lstm_layers=5,
        min_lstm_dims=10, max_lstm_dims=100):
    """ Generate a hyperparameter set that defines a DeepConvLSTM model."""
    hyperparameters = generate_base_hyper_parameter_set()
    number_of_conv_layers = np.random.randint(min_conv_layers, max_conv_layers)
    hyperparameters['filters'] = np.random.randint(
        min_conv_filters, max_conv_filters, number_of_conv_layers)
    number_of_lstm_layers = np.random.randint(min_lstm_layers, max_lstm_layers)
    hyperparameters['lstm_dims'] = np.random.randint(
        min_lstm_dims, max_lstm_dims, number_of_lstm_layers)
    return hyperparameters


def generate_base_hyper_parameter_set():

    """ Generate a base set of hyperparameters that are necessary for any model, but sufficient for none."""

    hyperparameters = {}
    hyperparameters['learning_rate'] = get_learning_rate()
    hyperparameters['regularization_rate'] = get_regularization()
    return hyperparameters


def get_learning_rate(low=1, high=4):
    """ Return random learning rate 10^-n where n is sampled uniformly between low and high bounds."""

    result = 10 ** (-np.random.uniform(low, high))
    return result


def get_regularization(low=1, high=4):
    """ Return random regularization rate 10^-n where n is sampled uniformly between low and high bounds."""

    return 10**(-np.random.uniform(low, high))


1	1	from keras.models import Sequential
2	1	from keras.layers import Dense, Activation, Convolution1D, Flatten, MaxPooling1D, Lambda, Convolution2D, Flatten, Reshape, LSTM, Dropout, TimeDistributed, Permute, BatchNormalization
		0 ignored issues – show Coding Style introduced 2016-07-07 14:47 UTC by Report Bug Copy Issue Report This line is too long as per the coding-style (182/79). This check looks for lines that are too long. You can specify the maximum line length. Loading history...
3	1	from keras.regularizers import l2
4	1	from keras.optimizers import Adam
5	1	import numpy as np
6
7
8	1	def generate_models(x_shape, number_of_classes, number_of_models=5, model_type=None, **kwargs):
		0 ignored issues – show Coding Style introduced 2016-07-07 14:47 UTC by Report Bug Copy Issue Report This line is too long as per the coding-style (95/79). This check looks for lines that are too long. You can specify the maximum line length. Loading history...
9		""" Generate one or multiple Keras models with random (default), or predefined, hyperparameters."""
		0 ignored issues – show Coding Style introduced 2016-07-07 14:47 UTC by Report Bug Copy Issue Report This line is too long as per the coding-style (103/79). This check looks for lines that are too long. You can specify the maximum line length. Loading history...
10	1	models = []
11	1	for _ in range(0, number_of_models):
12	1	if model_type is None: # random model choice:
13	1	current_model_type = 'CNN' if np.random.random(
14		) < 0.5 else 'DeepConvLSTM'
15		else: # user-defined model choice:
16		current_model_type = model_type
17
18	1	if current_model_type == 'CNN':
19	1	generate_model = generate_CNN_model # object is a function
20	1	generate_hyperparameter_set = generate_CNN_hyperparameter_set
21	1	if current_model_type == 'DeepConvLSTM':
22		generate_model = generate_DeepConvLSTM_model # object is a function
		0 ignored issues – show Coding Style introduced 2016-07-27 12:01 UTC by Report Bug Copy Issue Report This line is too long as per the coding-style (80/79). This check looks for lines that are too long. You can specify the maximum line length. Loading history...
23		generate_hyperparameter_set = generate_DeepConvLSTM_hyperparameter_set
		0 ignored issues – show Coding Style introduced 2016-07-07 14:47 UTC by Report Bug Copy Issue Report This line is too long as per the coding-style (82/79). This check looks for lines that are too long. You can specify the maximum line length. Loading history...
24	1	hyperparameters = generate_hyperparameter_set(**kwargs)
25	1	models.append(
26		(generate_model(x_shape, number_of_classes, **hyperparameters), hyperparameters, current_model_type))
		0 ignored issues – show Coding Style introduced 2016-07-07 14:47 UTC by Report Bug Copy Issue Report This line is too long as per the coding-style (113/79). This check looks for lines that are too long. You can specify the maximum line length. Loading history...
27	1	return models
28
29
30	1	def generate_DeepConvLSTM_model(x_shape, class_number, filters, lstm_dims, learning_rate=0.01,
		0 ignored issues – show Coding Style introduced 2016-07-07 14:47 UTC by Report Bug Copy Issue Report This line is too long as per the coding-style (94/79). This check looks for lines that are too long. You can specify the maximum line length. Loading history... Coding Style Naming introduced 2016-07-07 14:47 UTC by Report Bug Copy Issue Report The name `generate_DeepConvLSTM_model` does not conform to the function naming conventions (`[a-z_][a-z0-9_]{2,30}$`). This check looks for invalid names for a range of different identifiers. You can set regular expressions to which the identifiers must conform if the defaults do not match your requirements. If your project includes a Pylint configuration file, the settings contained in that file take precedence. To find out more about Pylint, please refer to their site. Loading history...
31		regularization_rate=0.01):
32		"""
33		Generate a model with convolution and LSTM layers.
34		See Ordonez et al., 2016, http://dx.doi.org/10.3390/s16010115
35
36		The compiled Keras model is returned.
37		"""
38	1	dim_length = x_shape[1] # number of samples in a time series
39	1	dim_channels = x_shape[2] # number of channels
40	1	output_dim = class_number # number of classes
41	1	weightinit = 'lecun_uniform' # weight initialization
42	1	model = Sequential() # initialize model
43	1	model.add(BatchNormalization(input_shape=(dim_length, dim_channels)))
44		# reshape a 2 dimensional array per file/person/object into a
45		# 3 dimensional array
46	1	model.add(
47		Reshape(target_shape=(1, dim_length, dim_channels)))
48	1	for filt in filters:
49		# filt: number of filters used in a layer
50		# filters: vector of filt values
51	1	model.add(
52		Convolution2D(filt, nb_row=3, nb_col=1, border_mode='same',
53		W_regularizer=l2(regularization_rate), init=weightinit))
		0 ignored issues – show Coding Style introduced 2016-07-27 11:53 UTC by Report Bug Copy Issue Report This line is too long as per the coding-style (82/79). This check looks for lines that are too long. You can specify the maximum line length. Loading history...
54	1	model.add(Activation('relu'))
55		# reshape 3 dimensional array back into a 2 dimensional array,
56		# but now with more dept as we have the the filters for each channel
57	1	model.add(Reshape(target_shape=(dim_length, filters[-1] * dim_channels)))
58
59	1	for lstm_dim in lstm_dims:
60	1	model.add(LSTM(output_dim=lstm_dim, return_sequences=True,
61		activation='tanh'))
62
63	1	model.add(Dropout(0.5)) # dropout before the dense layer
64		# set up final dense layer such that every timestamp is given one classification
		0 ignored issues – show Coding Style introduced 2016-07-07 14:47 UTC by Report Bug Copy Issue Report This line is too long as per the coding-style (84/79). This check looks for lines that are too long. You can specify the maximum line length. Loading history...
65	1	model.add(TimeDistributed(Dense(output_dim, W_regularizer=l2(regularization_rate))))
		0 ignored issues – show Coding Style introduced 2016-07-07 14:47 UTC by Report Bug Copy Issue Report This line is too long as per the coding-style (88/79). This check looks for lines that are too long. You can specify the maximum line length. Loading history...
66	1	model.add(Activation("softmax"))
67		# Final classification layer - per timestep
68	1	model.add(Lambda(lambda x: x[:, -1, :], output_shape=[output_dim]))
69
70	1	model.compile(loss='categorical_crossentropy',
71		optimizer=Adam(lr=learning_rate),
72		metrics=['accuracy'])
73
74	1	return model
75
76
77	1	def generate_CNN_model(x_shape, class_number, filters, fc_hidden_nodes,
		0 ignored issues – show Coding Style Naming introduced 2016-07-07 14:47 UTC by Report Bug Copy Issue Report The name `generate_CNN_model` does not conform to the function naming conventions (`[a-z_][a-z0-9_]{2,30}$`). This check looks for invalid names for a range of different identifiers. You can set regular expressions to which the identifiers must conform if the defaults do not match your requirements. If your project includes a Pylint configuration file, the settings contained in that file take precedence. To find out more about Pylint, please refer to their site. Loading history...
78		learning_rate=0.01, regularization_rate=0.01):
79		"""
80		Generate a convolutional neural network (CNN) model.
81
82		The compiled Keras model is returned.
83		"""
84	1	dim_length = x_shape[1] # number of samples in a time series
85	1	dim_channels = x_shape[2] # number of channels
86	1	outputdim = class_number # number of classes
87	1	weightinit = 'lecun_uniform' # weight initialization
88	1	model = Sequential()
89	1	model.add(BatchNormalization(input_shape=(dim_length, dim_channels)))
90	1	for filter_number in filters:
91	1	model.add(Convolution1D(filter_number, 3, border_mode='same',
92		W_regularizer=l2(regularization_rate), init=weightinit))
		0 ignored issues – show Coding Style introduced 2016-07-27 12:01 UTC by Report Bug Copy Issue Report This line is too long as per the coding-style (88/79). This check looks for lines that are too long. You can specify the maximum line length. Loading history...
93	1	model.add(Activation('relu'))
94	1	model.add(Flatten())
95	1	model.add(Dense(output_dim=fc_hidden_nodes,
96		W_regularizer=l2(regularization_rate), init=weightinit)) # Fully connected layer
		0 ignored issues – show Coding Style introduced 2016-07-20 14:18 UTC by Report Bug Copy Issue Report This line is too long as per the coding-style (101/79). This check looks for lines that are too long. You can specify the maximum line length. Loading history...
97	1	model.add(Activation('relu')) # Relu activation
98	1	model.add(Dense(output_dim=outputdim, init=weightinit))
99	1	model.add(Activation("softmax")) # Final classification layer
100
101	1	model.compile(loss='categorical_crossentropy',
102		optimizer=Adam(lr=learning_rate),
103		metrics=['accuracy'])
104
105	1	return model
106
107
108	1	def generate_CNN_hyperparameter_set(min_layers=1, max_layers=10,
		0 ignored issues – show Coding Style Naming introduced 2016-07-07 14:47 UTC by Report Bug Copy Issue Report The name `generate_CNN_hyperparameter_set` does not conform to the function naming conventions (`[a-z_][a-z0-9_]{2,30}$`). This check looks for invalid names for a range of different identifiers. You can set regular expressions to which the identifiers must conform if the defaults do not match your requirements. If your project includes a Pylint configuration file, the settings contained in that file take precedence. To find out more about Pylint, please refer to their site. Loading history...
109		min_filters=10, max_filters=100,
110		min_fc_nodes=10, max_fc_nodes=100):
111		""" Generate a hyperparameter set that define a CNN model."""
112	1	hyperparameters = generate_base_hyper_parameter_set()
113	1	number_of_layers = np.random.randint(min_layers, max_layers)
114	1	hyperparameters['filters'] = np.random.randint(
115		min_filters, max_filters, number_of_layers)
116	1	hyperparameters['fc_hidden_nodes'] = np.random.randint(
117		min_fc_nodes, max_fc_nodes)
118	1	return hyperparameters
119
120
121	1	def generate_DeepConvLSTM_hyperparameter_set(
		0 ignored issues – show Coding Style Naming introduced 2016-07-07 14:47 UTC by Report Bug Copy Issue Report The name `generate_DeepConvLSTM_hyperparameter_set` does not conform to the function naming conventions (`[a-z_][a-z0-9_]{2,30}$`). This check looks for invalid names for a range of different identifiers. You can set regular expressions to which the identifiers must conform if the defaults do not match your requirements. If your project includes a Pylint configuration file, the settings contained in that file take precedence. To find out more about Pylint, please refer to their site. Loading history...
122		min_conv_layers=1, max_conv_layers=10,
123		min_conv_filters=10, max_conv_filters=100,
124		min_lstm_layers=1, max_lstm_layers=5,
125		min_lstm_dims=10, max_lstm_dims=100):
126		""" Generate a hyperparameter set that defines a DeepConvLSTM model."""
127		hyperparameters = generate_base_hyper_parameter_set()
128		number_of_conv_layers = np.random.randint(min_conv_layers, max_conv_layers)
129		hyperparameters['filters'] = np.random.randint(
130		min_conv_filters, max_conv_filters, number_of_conv_layers)
131		number_of_lstm_layers = np.random.randint(min_lstm_layers, max_lstm_layers)
132		hyperparameters['lstm_dims'] = np.random.randint(
133		min_lstm_dims, max_lstm_dims, number_of_lstm_layers)
134		return hyperparameters
135
136
137	1	def generate_base_hyper_parameter_set():
		0 ignored issues – show Coding Style Naming introduced 2016-07-07 14:47 UTC by Report Bug Copy Issue Report The name `generate_base_hyper_parameter_set` does not conform to the function naming conventions (`[a-z_][a-z0-9_]{2,30}$`). This check looks for invalid names for a range of different identifiers. You can set regular expressions to which the identifiers must conform if the defaults do not match your requirements. If your project includes a Pylint configuration file, the settings contained in that file take precedence. To find out more about Pylint, please refer to their site. Loading history...
138		""" Generate a base set of hyperparameters that are necessary for any model, but sufficient for none."""
		0 ignored issues – show Coding Style introduced 2016-07-07 14:47 UTC by Report Bug Copy Issue Report This line is too long as per the coding-style (108/79). This check looks for lines that are too long. You can specify the maximum line length. Loading history...
139	1	hyperparameters = {}
140	1	hyperparameters['learning_rate'] = get_learning_rate()
141	1	hyperparameters['regularization_rate'] = get_regularization()
142	1	return hyperparameters
143
144
145	1	def get_learning_rate(low=1, high=4):
146		""" Return random learning rate 10^-n where n is sampled uniformly between low and high bounds."""
		0 ignored issues – show Coding Style introduced 2016-07-07 14:47 UTC by Report Bug Copy Issue Report This line is too long as per the coding-style (102/79). This check looks for lines that are too long. You can specify the maximum line length. Loading history...
147	1	result = 10 ** (-np.random.uniform(low, high))
148	1	return result
149
150
151	1	def get_regularization(low=1, high=4):
152		""" Return random regularization rate 10^-n where n is sampled uniformly between low and high bounds."""
		0 ignored issues – show Coding Style introduced 2016-07-07 14:47 UTC by Report Bug Copy Issue Report This line is too long as per the coding-style (108/79). This check looks for lines that are too long. You can specify the maximum line length. Loading history...
153		return 10**(-np.random.uniform(low, high))
154

NLeSC / mcfly

Push — master ( a63bb3...d33a20 )

generate_DeepConvLSTM_model() B

Complexity

Size

Duplication

Code Coverage

Importance

Duplication Side-by-Side

Filter issues like