generate_models() - Code Metrics - Inspection of "Merge pull request #147 from NLeSC/newmetrics" - NLeSC/mcfly - Measure and Improve Code Quality continuously with Scrutinizer

Completed

Push — master ( 3e0ec6...69dab7 )

by Christiaan

created 2017-09-18 08:47 UTC

generate_models() B

↳ Parent: Project

Complexity

Conditions

Size

Total Lines

108

Duplication

Lines	0
Ratio	0 %

Code Coverage

Tests	14
CRAP Score	6.0106

Importance

Changes	1
Bugs	0	Features	0

Metric	Value
cc	6
c	1
b	0
f	0
dl	0
loc	108
ccs	14
cts	15
cp	0.9333
crap	6.0106
rs	7.0769

How to fix Long Method

from keras.models import Sequential
from keras.layers import Dense, Activation, Convolution1D, Lambda, \
    Convolution2D, Flatten, \
    Reshape, LSTM, Dropout, TimeDistributed, 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, metrics=['accuracy'],
        model_type=None,
        cnn_min_layers=1, cnn_max_layers=10,
        cnn_min_filters=10, cnn_max_filters=100,
        cnn_min_fc_nodes=10, cnn_max_fc_nodes=2000,
        deepconvlstm_min_conv_layers=1, deepconvlstm_max_conv_layers=10,
        deepconvlstm_min_conv_filters=10, deepconvlstm_max_conv_filters=100,
        deepconvlstm_min_lstm_layers=1, deepconvlstm_max_lstm_layers=5,
        deepconvlstm_min_lstm_dims=10, deepconvlstm_max_lstm_dims=100,
        low_lr=1, high_lr=4, low_reg=1, high_reg=4
):
    """
    Generate one or multiple untrained Keras models with random hyperparameters.


    Parameters
    ----------
    x_shape : tuple
        Shape of the input dataset: (num_samples, num_timesteps, num_channels)
    number_of_classes : int
        Number of classes for classification task
    number_of_models : int
        Number of models to generate
    metrics : list
        Metrics to calculate on the validation set.
        See https://keras.io/metrics/ for possible values.
    model_type : str, optional
        Type of model to build: 'CNN' or 'DeepConvLSTM'.
        Default option None generates both models.
    cnn_min_layers : int
        minimum of Conv layers in CNN model
    cnn_max_layers : int
        maximum of Conv layers in CNN model
    cnn_min_filters : int
        minimum number of filters per Conv layer in CNN model
    cnn_max_filters : int
        maximum number of filters per Conv layer in CNN model
    cnn_min_fc_nodes : int
        minimum number of hidden nodes per Dense layer in CNN model
    cnn_max_fc_nodes : int
        maximum number of hidden nodes per Dense layer in CNN model
    deepconvlstm_min_conv_layers : int
        minimum number of Conv layers in DeepConvLSTM model
    deepconvlstm_max_conv_layers : int
        maximum number of Conv layers in DeepConvLSTM model
    deepconvlstm_min_conv_filters : int
        minimum number of filters per Conv layer in DeepConvLSTM model
    deepconvlstm_max_conv_filters : int
        maximum number of filters per Conv layer in DeepConvLSTM model
    deepconvlstm_min_lstm_layers : int
        minimum number of Conv layers in DeepConvLSTM model
    deepconvlstm_max_lstm_layers : int
        maximum number of Conv layers in DeepConvLSTM model
    deepconvlstm_min_lstm_dims : int
        minimum number of hidden nodes per LSTM layer in DeepConvLSTM model
    deepconvlstm_max_lstm_dims : int
        maximum number of hidden nodes per LSTM layer in DeepConvLSTM model
    low_lr : float
        minimum of log range for learning rate: learning rate is sampled
        between `10**(-low_reg)` and `10**(-high_reg)`
    high_lr : float
        maximum  of log range for learning rate: learning rate is sampled
        between `10**(-low_reg)` and `10**(-high_reg)`
    low_reg : float
        minimum  of log range for regularization rate: regularization rate is
        sampled between `10**(-low_reg)` and `10**(-high_reg)`
    high_reg : float
        maximum  of log range for regularization rate: regularization rate is
        sampled between `10**(-low_reg)` and `10**(-high_reg)`

    Returns
    -------
    models : list
        List of compiled models
    """
    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
        generate_model = None
        if current_model_type == 'CNN':
            generate_model = generate_CNN_model  # generate_model is a function
            hyperparameters = generate_CNN_hyperparameter_set(
                min_layers=cnn_min_layers, max_layers=cnn_max_layers,
                min_filters=cnn_min_filters, max_filters=cnn_max_filters,
                min_fc_nodes=cnn_min_fc_nodes, max_fc_nodes=cnn_max_fc_nodes,
                low_lr=low_lr, high_lr=high_lr, low_reg=low_reg,
                high_reg=high_reg)
        if current_model_type == 'DeepConvLSTM':
            generate_model = generate_DeepConvLSTM_model
            hyperparameters = generate_DeepConvLSTM_hyperparameter_set(
                min_conv_layers=deepconvlstm_min_conv_layers,
                max_conv_layers=deepconvlstm_max_conv_layers,
                min_conv_filters=deepconvlstm_min_conv_filters,
                max_conv_filters=deepconvlstm_max_conv_filters,
                min_lstm_layers=deepconvlstm_min_lstm_layers,
                max_lstm_layers=deepconvlstm_max_lstm_layers,
                min_lstm_dims=deepconvlstm_min_lstm_dims,
                max_lstm_dims=deepconvlstm_max_lstm_dims,
                low_lr=low_lr, high_lr=high_lr, low_reg=low_reg,
                high_reg=high_reg)
        models.append(
            (generate_model(x_shape, number_of_classes, metrics=metrics, **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, metrics=['accuracy']):
    """
    Generate a model with convolution and LSTM layers.
    See Ordonez et al., 2016, http://dx.doi.org/10.3390/s16010115

    Parameters
    ----------
    x_shape : tuple
        Shape of the input dataset: (num_samples, num_timesteps, num_channels)
    class_number : int
        Number of classes for classification task
    filters : list of ints
        number of filters for each convolutional layer
    lstm_dims : list of ints
        number of hidden nodes for each LSTM layer
    learning_rate : float
        learning rate
    regularization_rate : float
        regularization rate
    metrics : list
        Metrics to calculate on the validation set.
        See https://keras.io/metrics/ for possible values.

    Returns
    -------
    model : Keras model
        The compiled Keras model
    """
    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=(dim_length, dim_channels, 1)))
    for filt in filters:
        # filt: number of filters used in a layer
        # filters: vector of filt values
        model.add(
            Convolution2D(filt, kernel_size=(3, 1), padding='same',
                          kernel_regularizer=l2(regularization_rate),
                          kernel_initializer=weightinit))
        model.add(BatchNormalization())
        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(units=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(units=output_dim, kernel_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=metrics)

    return model


def generate_CNN_model(x_shape, class_number, filters, fc_hidden_nodes,
                       learning_rate=0.01, regularization_rate=0.01,
                       metrics=['accuracy']):
    """
    Generate a convolutional neural network (CNN) model.

    The compiled Keras model is returned.

    Parameters
    ----------
    x_shape : tuple
        Shape of the input dataset: (num_samples, num_timesteps, num_channels)
    class_number : int
        Number of classes for classification task
    filters : list of ints
        number of filters for each convolutional layer
    fc_hidden_nodes : int
        number of hidden nodes for the hidden dense layer
    learning_rate : float
        learning rate
    regularization_rate : float
        regularization rate
    metrics : list
        Metrics to calculate on the validation set.
        See https://keras.io/metrics/ for possible values.

    Returns
    -------
    model : Keras model
        The compiled Keras model
    """
    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, kernel_size=3, padding='same',
                                kernel_regularizer=l2(regularization_rate),
                                kernel_initializer=weightinit))
        model.add(BatchNormalization())
        model.add(Activation('relu'))
    model.add(Flatten())
    model.add(Dense(units=fc_hidden_nodes,
                    kernel_regularizer=l2(regularization_rate),
                    kernel_initializer=weightinit))  # Fully connected layer
    model.add(Activation('relu'))  # Relu activation
    model.add(Dense(units=outputdim, kernel_initializer=weightinit))
    model.add(BatchNormalization())
    model.add(Activation("softmax"))  # Final classification layer

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

    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=2000,
                                    low_lr=1, high_lr=4, low_reg=1,
                                    high_reg=4):
    """ Generate a hyperparameter set that define a CNN model.

    Parameters
    ----------
    min_layers : int
        minimum of Conv layers
    max_layers : int
        maximum of Conv layers
    min_filters : int
        minimum number of filters per Conv layer
    max_filters : int
        maximum number of filters per Conv layer
    min_fc_nodes : int
        minimum number of hidden nodes per Dense layer
    max_fc_nodes : int
        maximum number of hidden nodes per Dense layer
    low_lr : float
        minimum of log range for learning rate: learning rate is sampled
        between `10**(-low_reg)` and `10**(-high_reg)`
    high_lr : float
        maximum  of log range for learning rate: learning rate is sampled
        between `10**(-low_reg)` and `10**(-high_reg)`
    low_reg : float
        minimum  of log range for regularization rate: regularization rate is
        sampled between `10**(-low_reg)` and `10**(-high_reg)`
    high_reg : float
        maximum  of log range for regularization rate: regularization rate is
        sampled between `10**(-low_reg)` and `10**(-high_reg)`

    Returns
    ----------
    hyperparameters : dict
        parameters for a CNN model
    """
    hyperparameters = generate_base_hyper_parameter_set(
        low_lr, high_lr, low_reg, high_reg)
    number_of_layers = np.random.randint(min_layers, max_layers + 1)
    hyperparameters['filters'] = np.random.randint(
        min_filters, max_filters + 1, number_of_layers)
    hyperparameters['fc_hidden_nodes'] = np.random.randint(
        min_fc_nodes, max_fc_nodes + 1)
    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,
        low_lr=1, high_lr=4, low_reg=1, high_reg=4):
    """ Generate a hyperparameter set that defines a DeepConvLSTM model.

    Parameters
    ----------
    min_conv_layers : int
        minimum number of Conv layers in DeepConvLSTM model
    max_conv_layers : int
        maximum number of Conv layers in DeepConvLSTM model
    min_conv_filters : int
        minimum number of filters per Conv layer in DeepConvLSTM model
    max_conv_filters : int
        maximum number of filters per Conv layer in DeepConvLSTM model
    min_lstm_layers : int
        minimum number of Conv layers in DeepConvLSTM model
    max_lstm_layers : int
        maximum number of Conv layers in DeepConvLSTM model
    min_lstm_dims : int
        minimum number of hidden nodes per LSTM layer in DeepConvLSTM model
    max_lstm_dims : int
        maximum number of hidden nodes per LSTM layer in DeepConvLSTM model
    low_lr : float
        minimum of log range for learning rate: learning rate is sampled
        between `10**(-low_reg)` and `10**(-high_reg)`
    high_lr : float
        maximum  of log range for learning rate: learning rate is sampled
        between `10**(-low_reg)` and `10**(-high_reg)`
    low_reg : float
        minimum  of log range for regularization rate: regularization rate is
        sampled between `10**(-low_reg)` and `10**(-high_reg)`
    high_reg : float
        maximum  of log range for regularization rate: regularization rate is
        sampled between `10**(-low_reg)` and `10**(-high_reg)`

    Returns
    ----------
    hyperparameters: dict
        hyperparameters for a DeepConvLSTM model
    """
    hyperparameters = generate_base_hyper_parameter_set(
        low_lr, high_lr, low_reg, high_reg)
    number_of_conv_layers = np.random.randint(
        min_conv_layers, max_conv_layers + 1)
    hyperparameters['filters'] = np.random.randint(
        min_conv_filters, max_conv_filters + 1, number_of_conv_layers).tolist()
    number_of_lstm_layers = np.random.randint(
        min_lstm_layers, max_lstm_layers + 1)
    hyperparameters['lstm_dims'] = np.random.randint(
        min_lstm_dims, max_lstm_dims + 1, number_of_lstm_layers).tolist()
    return hyperparameters


def generate_base_hyper_parameter_set(
        low_lr=1,
        high_lr=4,
        low_reg=1,
        high_reg=4):
    """ Generate a base set of hyperparameters that are necessary for any
    model, but sufficient for none.

    Parameters
    ----------
    low_lr : float
        minimum of log range for learning rate: learning rate is sampled
        between `10**(-low_reg)` and `10**(-high_reg)`
    high_lr : float
        maximum  of log range for learning rate: learning rate is sampled
        between `10**(-low_reg)` and `10**(-high_reg)`
    low_reg : float
        minimum  of log range for regularization rate: regularization rate is
        sampled between `10**(-low_reg)` and `10**(-high_reg)`
    high_reg : float
        maximum  of log range for regularization rate: regularization rate is
        sampled between `10**(-low_reg)` and `10**(-high_reg)`

    Returns
    -------
    hyperparameters : dict
        basis hyperpameters
    """
    hyperparameters = {}
    hyperparameters['learning_rate'] = get_learning_rate(low_lr, high_lr)
    hyperparameters['regularization_rate'] = get_regularization(
        low_reg, high_reg)
    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.

    Parameters
    ----------
    low : float
        low bound
    high : float
        high bound

    Returns
    -------
    learning_rate : float
        learning rate
    """
    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.

    Parameters
    ----------
    low : float
        low bound
    high : float
        high bound

    Returns
    -------
    regularization_rate : float
        regularization rate
    """
    return 10 ** (-np.random.uniform(low, high))


1	1	from keras.models import Sequential
2	1	from keras.layers import Dense, Activation, Convolution1D, Lambda, \
3		Convolution2D, Flatten, \
4		Reshape, LSTM, Dropout, TimeDistributed, BatchNormalization
5	1	from keras.regularizers import l2
6	1	from keras.optimizers import Adam
7	1	import numpy as np
8
9
10	1	def generate_models(
11		x_shape, number_of_classes, number_of_models=5, metrics=['accuracy'],
12		model_type=None,
13		cnn_min_layers=1, cnn_max_layers=10,
14		cnn_min_filters=10, cnn_max_filters=100,
15		cnn_min_fc_nodes=10, cnn_max_fc_nodes=2000,
16		deepconvlstm_min_conv_layers=1, deepconvlstm_max_conv_layers=10,
17		deepconvlstm_min_conv_filters=10, deepconvlstm_max_conv_filters=100,
18		deepconvlstm_min_lstm_layers=1, deepconvlstm_max_lstm_layers=5,
19		deepconvlstm_min_lstm_dims=10, deepconvlstm_max_lstm_dims=100,
20		low_lr=1, high_lr=4, low_reg=1, high_reg=4
21		):
22		"""
23		Generate one or multiple untrained Keras models with random hyperparameters.
		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...
24
25		Parameters
26		----------
27		x_shape : tuple
28		Shape of the input dataset: (num_samples, num_timesteps, num_channels)
29		number_of_classes : int
30		Number of classes for classification task
31		number_of_models : int
32		Number of models to generate
33		metrics : list
34		Metrics to calculate on the validation set.
35		See https://keras.io/metrics/ for possible values.
36		model_type : str, optional
37		Type of model to build: 'CNN' or 'DeepConvLSTM'.
38		Default option None generates both models.
39		cnn_min_layers : int
40		minimum of Conv layers in CNN model
41		cnn_max_layers : int
42		maximum of Conv layers in CNN model
43		cnn_min_filters : int
44		minimum number of filters per Conv layer in CNN model
45		cnn_max_filters : int
46		maximum number of filters per Conv layer in CNN model
47		cnn_min_fc_nodes : int
48		minimum number of hidden nodes per Dense layer in CNN model
49		cnn_max_fc_nodes : int
50		maximum number of hidden nodes per Dense layer in CNN model
51		deepconvlstm_min_conv_layers : int
52		minimum number of Conv layers in DeepConvLSTM model
53		deepconvlstm_max_conv_layers : int
54		maximum number of Conv layers in DeepConvLSTM model
55		deepconvlstm_min_conv_filters : int
56		minimum number of filters per Conv layer in DeepConvLSTM model
57		deepconvlstm_max_conv_filters : int
58		maximum number of filters per Conv layer in DeepConvLSTM model
59		deepconvlstm_min_lstm_layers : int
60		minimum number of Conv layers in DeepConvLSTM model
61		deepconvlstm_max_lstm_layers : int
62		maximum number of Conv layers in DeepConvLSTM model
63		deepconvlstm_min_lstm_dims : int
64		minimum number of hidden nodes per LSTM layer in DeepConvLSTM model
65		deepconvlstm_max_lstm_dims : int
66		maximum number of hidden nodes per LSTM layer in DeepConvLSTM model
67		low_lr : float
68		minimum of log range for learning rate: learning rate is sampled
69		between `10(-low_reg)` and `10(-high_reg)`
70		high_lr : float
71		maximum of log range for learning rate: learning rate is sampled
72		between `10(-low_reg)` and `10(-high_reg)`
73		low_reg : float
74		minimum of log range for regularization rate: regularization rate is
75		sampled between `10(-low_reg)` and `10(-high_reg)`
76		high_reg : float
77		maximum of log range for regularization rate: regularization rate is
78		sampled between `10(-low_reg)` and `10(-high_reg)`
79
80		Returns
81		-------
82		models : list
83		List of compiled models
84		"""
85	1	models = []
86	1	for _ in range(0, number_of_models):
87	1	if model_type is None: # random model choice:
88	1	current_model_type = 'CNN' if np.random.random(
89		) < 0.5 else 'DeepConvLSTM'
90		else: # user-defined model choice:
91		current_model_type = model_type
92	1	generate_model = None
93	1	if current_model_type == 'CNN':
94	1	generate_model = generate_CNN_model # generate_model is a function
95	1	hyperparameters = generate_CNN_hyperparameter_set(
96		min_layers=cnn_min_layers, max_layers=cnn_max_layers,
97		min_filters=cnn_min_filters, max_filters=cnn_max_filters,
98		min_fc_nodes=cnn_min_fc_nodes, max_fc_nodes=cnn_max_fc_nodes,
99		low_lr=low_lr, high_lr=high_lr, low_reg=low_reg,
100		high_reg=high_reg)
101	1	if current_model_type == 'DeepConvLSTM':
102	1	generate_model = generate_DeepConvLSTM_model
103	1	hyperparameters = generate_DeepConvLSTM_hyperparameter_set(
104		min_conv_layers=deepconvlstm_min_conv_layers,
105		max_conv_layers=deepconvlstm_max_conv_layers,
106		min_conv_filters=deepconvlstm_min_conv_filters,
107		max_conv_filters=deepconvlstm_max_conv_filters,
108		min_lstm_layers=deepconvlstm_min_lstm_layers,
109		max_lstm_layers=deepconvlstm_max_lstm_layers,
110		min_lstm_dims=deepconvlstm_min_lstm_dims,
111		max_lstm_dims=deepconvlstm_max_lstm_dims,
112		low_lr=low_lr, high_lr=high_lr, low_reg=low_reg,
113		high_reg=high_reg)
114	1	models.append(
115		(generate_model(x_shape, number_of_classes, metrics=metrics, **hyperparameters),
		0 ignored issues – show Coding Style introduced 2017-06-29 15:18 UTC by Report Bug Copy Issue Report This line is too long as per the coding-style (92/79). This check looks for lines that are too long. You can specify the maximum line length. Loading history...
116		hyperparameters, current_model_type))
117	1	return models
118
119
120	1	def generate_DeepConvLSTM_model(
121		x_shape, class_number, filters, lstm_dims, learning_rate=0.01,
122		regularization_rate=0.01, metrics=['accuracy']):
123		"""
124		Generate a model with convolution and LSTM layers.
125		See Ordonez et al., 2016, http://dx.doi.org/10.3390/s16010115
126
127		Parameters
128		----------
129		x_shape : tuple
130		Shape of the input dataset: (num_samples, num_timesteps, num_channels)
131		class_number : int
132		Number of classes for classification task
133		filters : list of ints
134		number of filters for each convolutional layer
135		lstm_dims : list of ints
136		number of hidden nodes for each LSTM layer
137		learning_rate : float
138		learning rate
139		regularization_rate : float
140		regularization rate
141		metrics : list
142		Metrics to calculate on the validation set.
143		See https://keras.io/metrics/ for possible values.
144
145		Returns
146		-------
147		model : Keras model
148		The compiled Keras model
149		"""
150	1	dim_length = x_shape[1] # number of samples in a time series
151	1	dim_channels = x_shape[2] # number of channels
152	1	output_dim = class_number # number of classes
153	1	weightinit = 'lecun_uniform' # weight initialization
154	1	model = Sequential() # initialize model
155	1	model.add(BatchNormalization(input_shape=(dim_length, dim_channels)))
156		# reshape a 2 dimensional array per file/person/object into a
157		# 3 dimensional array
158	1	model.add(
159		Reshape(target_shape=(dim_length, dim_channels, 1)))
160	1	for filt in filters:
161		# filt: number of filters used in a layer
162		# filters: vector of filt values
163	1	model.add(
164		Convolution2D(filt, kernel_size=(3, 1), padding='same',
165		kernel_regularizer=l2(regularization_rate),
166		kernel_initializer=weightinit))
167	1	model.add(BatchNormalization())
168	1	model.add(Activation('relu'))
169		# reshape 3 dimensional array back into a 2 dimensional array,
170		# but now with more dept as we have the the filters for each channel
171	1	model.add(Reshape(target_shape=(dim_length, filters[-1] * dim_channels)))
172
173	1	for lstm_dim in lstm_dims:
174	1	model.add(LSTM(units=lstm_dim, return_sequences=True,
175		activation='tanh'))
176
177	1	model.add(Dropout(0.5)) # dropout before the dense layer
178		# set up final dense layer such that every timestamp is given one
179		# classification
180	1	model.add(
181		TimeDistributed(
182		Dense(units=output_dim, kernel_regularizer=l2(regularization_rate))))
		0 ignored issues – show Coding Style introduced 2017-04-04 11:41 UTC by Report Bug Copy Issue Report This line is too long as per the coding-style (81/79). This check looks for lines that are too long. You can specify the maximum line length. Loading history...
183	1	model.add(Activation("softmax"))
184		# Final classification layer - per timestep
185	1	model.add(Lambda(lambda x: x[:, -1, :], output_shape=[output_dim]))
186
187	1	model.compile(loss='categorical_crossentropy',
188		optimizer=Adam(lr=learning_rate),
189		metrics=metrics)
190
191	1	return model
192
193
194	1	def generate_CNN_model(x_shape, class_number, filters, fc_hidden_nodes,
195		learning_rate=0.01, regularization_rate=0.01,
196		metrics=['accuracy']):
197		"""
198		Generate a convolutional neural network (CNN) model.
199
200		The compiled Keras model is returned.
201
202		Parameters
203		----------
204		x_shape : tuple
205		Shape of the input dataset: (num_samples, num_timesteps, num_channels)
206		class_number : int
207		Number of classes for classification task
208		filters : list of ints
209		number of filters for each convolutional layer
210		fc_hidden_nodes : int
211		number of hidden nodes for the hidden dense layer
212		learning_rate : float
213		learning rate
214		regularization_rate : float
215		regularization rate
216		metrics : list
217		Metrics to calculate on the validation set.
218		See https://keras.io/metrics/ for possible values.
219
220		Returns
221		-------
222		model : Keras model
223		The compiled Keras model
224		"""
225	1	dim_length = x_shape[1] # number of samples in a time series
226	1	dim_channels = x_shape[2] # number of channels
227	1	outputdim = class_number # number of classes
228	1	weightinit = 'lecun_uniform' # weight initialization
229	1	model = Sequential()
230	1	model.add(
231		BatchNormalization(
232		input_shape=(
233		dim_length,
234		dim_channels)))
235	1	for filter_number in filters:
236	1	model.add(Convolution1D(filter_number, kernel_size=3, padding='same',
237		kernel_regularizer=l2(regularization_rate),
238		kernel_initializer=weightinit))
239	1	model.add(BatchNormalization())
240	1	model.add(Activation('relu'))
241	1	model.add(Flatten())
242	1	model.add(Dense(units=fc_hidden_nodes,
243		kernel_regularizer=l2(regularization_rate),
244		kernel_initializer=weightinit)) # Fully connected layer
245	1	model.add(Activation('relu')) # Relu activation
246	1	model.add(Dense(units=outputdim, kernel_initializer=weightinit))
247	1	model.add(BatchNormalization())
248	1	model.add(Activation("softmax")) # Final classification layer
249
250	1	model.compile(loss='categorical_crossentropy',
251		optimizer=Adam(lr=learning_rate),
252		metrics=metrics)
253
254	1	return model
255
256
257	1	def generate_CNN_hyperparameter_set(min_layers=1, max_layers=10,
258		min_filters=10, max_filters=100,
259		min_fc_nodes=10, max_fc_nodes=2000,
260		low_lr=1, high_lr=4, low_reg=1,
261		high_reg=4):
262		""" Generate a hyperparameter set that define a CNN model.
263
264		Parameters
265		----------
266		min_layers : int
267		minimum of Conv layers
268		max_layers : int
269		maximum of Conv layers
270		min_filters : int
271		minimum number of filters per Conv layer
272		max_filters : int
273		maximum number of filters per Conv layer
274		min_fc_nodes : int
275		minimum number of hidden nodes per Dense layer
276		max_fc_nodes : int
277		maximum number of hidden nodes per Dense layer
278		low_lr : float
279		minimum of log range for learning rate: learning rate is sampled
280		between `10(-low_reg)` and `10(-high_reg)`
281		high_lr : float
282		maximum of log range for learning rate: learning rate is sampled
283		between `10(-low_reg)` and `10(-high_reg)`
284		low_reg : float
285		minimum of log range for regularization rate: regularization rate is
286		sampled between `10(-low_reg)` and `10(-high_reg)`
287		high_reg : float
288		maximum of log range for regularization rate: regularization rate is
289		sampled between `10(-low_reg)` and `10(-high_reg)`
290
291		Returns
292		----------
293		hyperparameters : dict
294		parameters for a CNN model
295		"""
296	1	hyperparameters = generate_base_hyper_parameter_set(
297		low_lr, high_lr, low_reg, high_reg)
298	1	number_of_layers = np.random.randint(min_layers, max_layers + 1)
299	1	hyperparameters['filters'] = np.random.randint(
300		min_filters, max_filters + 1, number_of_layers)
301	1	hyperparameters['fc_hidden_nodes'] = np.random.randint(
302		min_fc_nodes, max_fc_nodes + 1)
303	1	return hyperparameters
304
305
306	1	def generate_DeepConvLSTM_hyperparameter_set(
307		min_conv_layers=1, max_conv_layers=10,
308		min_conv_filters=10, max_conv_filters=100,
309		min_lstm_layers=1, max_lstm_layers=5,
310		min_lstm_dims=10, max_lstm_dims=100,
311		low_lr=1, high_lr=4, low_reg=1, high_reg=4):
312		""" Generate a hyperparameter set that defines a DeepConvLSTM model.
313
314		Parameters
315		----------
316		min_conv_layers : int
317		minimum number of Conv layers in DeepConvLSTM model
318		max_conv_layers : int
319		maximum number of Conv layers in DeepConvLSTM model
320		min_conv_filters : int
321		minimum number of filters per Conv layer in DeepConvLSTM model
322		max_conv_filters : int
323		maximum number of filters per Conv layer in DeepConvLSTM model
324		min_lstm_layers : int
325		minimum number of Conv layers in DeepConvLSTM model
326		max_lstm_layers : int
327		maximum number of Conv layers in DeepConvLSTM model
328		min_lstm_dims : int
329		minimum number of hidden nodes per LSTM layer in DeepConvLSTM model
330		max_lstm_dims : int
331		maximum number of hidden nodes per LSTM layer in DeepConvLSTM model
332		low_lr : float
333		minimum of log range for learning rate: learning rate is sampled
334		between `10(-low_reg)` and `10(-high_reg)`
335		high_lr : float
336		maximum of log range for learning rate: learning rate is sampled
337		between `10(-low_reg)` and `10(-high_reg)`
338		low_reg : float
339		minimum of log range for regularization rate: regularization rate is
340		sampled between `10(-low_reg)` and `10(-high_reg)`
341		high_reg : float
342		maximum of log range for regularization rate: regularization rate is
343		sampled between `10(-low_reg)` and `10(-high_reg)`
344
345		Returns
346		----------
347		hyperparameters: dict
348		hyperparameters for a DeepConvLSTM model
349		"""
350	1	hyperparameters = generate_base_hyper_parameter_set(
351		low_lr, high_lr, low_reg, high_reg)
352	1	number_of_conv_layers = np.random.randint(
353		min_conv_layers, max_conv_layers + 1)
354	1	hyperparameters['filters'] = np.random.randint(
355		min_conv_filters, max_conv_filters + 1, number_of_conv_layers).tolist()
356	1	number_of_lstm_layers = np.random.randint(
357		min_lstm_layers, max_lstm_layers + 1)
358	1	hyperparameters['lstm_dims'] = np.random.randint(
359		min_lstm_dims, max_lstm_dims + 1, number_of_lstm_layers).tolist()
360	1	return hyperparameters
361
362
363	1	def generate_base_hyper_parameter_set(
364		low_lr=1,
365		high_lr=4,
366		low_reg=1,
367		high_reg=4):
368		""" Generate a base set of hyperparameters that are necessary for any
369		model, but sufficient for none.
370
371		Parameters
372		----------
373		low_lr : float
374		minimum of log range for learning rate: learning rate is sampled
375		between `10(-low_reg)` and `10(-high_reg)`
376		high_lr : float
377		maximum of log range for learning rate: learning rate is sampled
378		between `10(-low_reg)` and `10(-high_reg)`
379		low_reg : float
380		minimum of log range for regularization rate: regularization rate is
381		sampled between `10(-low_reg)` and `10(-high_reg)`
382		high_reg : float
383		maximum of log range for regularization rate: regularization rate is
384		sampled between `10(-low_reg)` and `10(-high_reg)`
385
386		Returns
387		-------
388		hyperparameters : dict
389		basis hyperpameters
390		"""
391	1	hyperparameters = {}
392	1	hyperparameters['learning_rate'] = get_learning_rate(low_lr, high_lr)
393	1	hyperparameters['regularization_rate'] = get_regularization(
394		low_reg, high_reg)
395	1	return hyperparameters
396
397
398	1	def get_learning_rate(low=1, high=4):
399		""" Return random learning rate 10^-n where n is sampled uniformly between
400		low and high bounds.
401
402		Parameters
403		----------
404		low : float
405		low bound
406		high : float
407		high bound
408
409		Returns
410		-------
411		learning_rate : float
412		learning rate
413		"""
414	1	result = 10 ** (-np.random.uniform(low, high))
415	1	return result
416
417
418	1	def get_regularization(low=1, high=4):
419		""" Return random regularization rate 10^-n where n is sampled uniformly
420		between low and high bounds.
421
422		Parameters
423		----------
424		low : float
425		low bound
426		high : float
427		high bound
428
429		Returns
430		-------
431		regularization_rate : float
432		regularization rate
433		"""
434		return 10 ** (-np.random.uniform(low, high))
435

NLeSC / mcfly

Push — master ( 3e0ec6...69dab7 )

generate_models() B

Complexity

Size

Duplication

Code Coverage

Importance

How to fix Long Method

Long Method

Duplication Side-by-Side

Filter issues like