skkeras.models module#

Models to be used with the build function.

class skkeras.models.Straight(convolution_filters=None, convolution_kernel_size=None, convolution_strides=None, convolution_padding='valid', convolution_dilation_rate=None, convolution_activation=None, convolution_use_bias=True, convolution_kernel_initializer='glorot_uniform', convolution_bias_initializer='zeros', convolution_kernel_regularizer_l1=None, convolution_kernel_regularizer_l2=None, convolution_bias_regularizer_l1=None, convolution_bias_regularizer_l2=None, convolution_activity_regularizer_l1=None, convolution_activity_regularizer_l2=None, convolution_kernel_constraint=None, convolution_bias_constraint=None, pooling_type='max', pooling_pool_size=None, pooling_strides=None, pooling_padding='valid', recurrent_type='lstm', recurrent_units=None, recurrent_activation='tanh', recurrent_recurrent_activation='hard_sigmoid', recurrent_use_bias=True, recurrent_kernel_initializer='glorot_uniform', recurrent_recurrent_initializer='orthogonal', recurrent_bias_initializer='zeros', recurrent_unit_forget_bias=True, recurrent_kernel_regularizer_l1=None, recurrent_kernel_regularizer_l2=None, recurrent_bias_regularizer_l1=None, recurrent_bias_regularizer_l2=None, recurrent_activity_regularizer_l1=None, recurrent_activity_regularizer_l2=None, recurrent_kernel_constraint=None, recurrent_recurrent_constraint=None, recurrent_bias_constraint=None, recurrent_dropout=0.0, recurrent_recurrent_dropout=0.0, recurrent_go_backwards=False, recurrent_stateful=False, recurrent_unroll=False, recurrent_implementation=1, batchnormalization=False, batchnormalization_axis=-1, batchnormalization_momentum=0.99, batchnormalization_epsilon=0.001, batchnormalization_center=True, batchnormalization_scale=True, batchnormalization_beta_initializer='zeros', batchnormalization_gamma_initializer='ones', batchnormalization_moving_mean_initializer='zeros', batchnormalization_moving_variance_initializer='ones', batchnormalization_beta_constraint=None, batchnormalization_gamma_constraint=None, dense_units=None, dense_activation='relu', dense_use_bias=True, dense_kernel_initializer='he_uniform', dense_bias_initializer='zeros', recurrent_regularizer_l1=None, recurrent_regularizer_l2=None, beta_regularizer_l1=None, beta_regularizer_l2=None, gamma_regularizer_l1=None, gamma_regularizer_l2=None, dense_kernel_regularizer_l1=None, dense_kernel_regularizer_l2=None, dense_bias_regularizer_l1=None, dense_bias_regularizer_l2=None, dense_activity_regularizer_l1=None, dense_activity_regularizer_l2=None, dense_kernel_constraint=None, dense_bias_constraint=None, dropout_rate=0.0, dropout_noise_shape=None, dropout_seed=None, return_sequences=False)[source]#

Bases: object

Straight feed-forward hidden-layers.

Basic straight feed-forward hidden-layer architecture.

Parameters:

convolution_filters (integer, default=None) – Dimensionality of the output space.
convolution_kernel_size (integer/tuple/list, default=None) – Dimensionality of the convolution window.
convolution_strides (integer/tuple/list, default=None) – Strides of the convolution.
convolution_padding ({"valid", "same"}, default='valid')
convolution_dilation_rate (integer/tuple/list, default=None) – Dilation rate to use for dilated convolution.
convolution_activation (string/function, default=None) – Activation function.
convolution_use_bias (boolean, default=True) – Whether the layer uses a bias vector.
convolution_kernel_initializer (string/function, default='glorot_uniform') – Initializer for the kernel weights matrix.
convolution_bias_initializer (string/function, default='zeros') – Initializer for the bias vector.
covolution_kernel_regularizer_l1 (float, default=None) – L1 regularization factor applied to the kernel weights matrix.
convolution_kernel_regularizer_l2 (float, default=None) – L2 regularization factor applied to the kernel weights matrix.
convolution_bias_regularizer_l1 (float, default=None) – L1 regularization factor applied to the bias vector.
convolution_bias_regularizer_l2 (float, default=None) – L2 regularization factor applied to the bias vector.
convolution_activity_regularizer_l1 (float, default=None) – L1 regularization factor applied to the output of the layer.
convolution_activity_regularizer_l2 (float, default=None) – L2 regularization factor applied to the output of the layer.
convolution_kernel_constraint (function, default=None) – Constraint function applied to the kernel matrix.
convolution_bias_constraint (function, default=None) – Constraint function applied to the bias vector.
pooling_type ({"max", "average}, default='max')
pooling_pool_size (integer/tuple/list, default=None) – Factors by which to downscale.
pooling_strides (integer/tuple/list, default=None) – Strides values.
pooling_padding ({"valid", "same"}, default='valid')
recurrent_type ({"lstm", "gru"}, default='lstm')
recurrent_units (integer, default=None) – Dimensionality of the output space.
recurrent_activation (string/function, default='tanh') – Activation function to use.
recurrent_recurrent_activation (string/function, default='hard_sigmoid') – Activation function to use for the recurrent step.
recurrent_use_bias (boolean, default=True) – Whether the layer uses a bias vector.
recurrent_kernel_initializer (string/function, default='glorot_uniform') – Initializer for the kernel weights matrix.
recurrent_recurrent_initializer (string/function, default='orthogonal') – Initializer for the recurrent_kernel weights matrix.
recurrent_bias_initializer (string/function, default='zeros') – Initializer for the bias vector.
recurrent_unit_forget_bias (boolean, default=True) – If True, add 1 to the bias of the forget gate at initialization.
recurrent_kernel_regularizer_l1 (float, default=None) – L1 regularization factor applied to the kernel weights matrix.
recurrent_kernel_regularizer_l2 (float, default=None) – L2 regularization factor applied to the kernel weights matrix.
recurrent_bias_regularizer_l1 (float, default=None) – L1 regularization factor applied to the bias vector.
recurrent_bias_regularizer_l2 (float, default=None) – L2 regularization factor applied to the bias vector.
recurrent_activity_regularizer_l1 (float, default=None) – L1 regularization factor applied to the output of the layer.
recurrent_activity_regularizer_l2 (float, default=None) – L2 regularization factor applied to the output of the layer.
recurrent_kernel_constraint (function, default=None) – Constraint function applied to the kernel weights matrix.
recurrent_recurrent_constraint (function, default=None) – Constraint function applied to the recurrent_kernel weights matrix.
recurrent_bias_constraint (function, default=None) – Constraint function applied to the bias vector.
recurrent_dropout (float in [0, 1], default=0.0) – Fraction of the units to drop for the linear transformation of the inputs.
recurrent_recurrent_dropout (float in [0, 1], default=0.0) – Fraction of the units to drop for the linear transformation of the recurrent state.
recurrent_go_backwards (boolean, default=False) – If True, process the input sequence backwards and return the reversed sequence.
recurrent_stateful (boolean, default=False) – If True, the last state for each sample at index i in a batch will be used as initial state for the sample of index i in the following batch.
recurrent_unroll (boolean, default=False) – If True, the network will be unrolled, else a symbolic loop will be used.
recurrent_implementation ({0, 1, 2}, default=1)
batchnormalization (boolean, default=False) – Whether to perform batch normalization or not.
batchnormalization_axis (integer, default=-1) – The axis that should be normalized (typically the features axis).
batchnormalization_momentum (float, default=0.99) – Momentum for the moving average.
batchnormalization_epsilon (float, default=0.001) – Small float added to variance to avoid dividing by zero.
batchnormalization_center (boolean, default=True) – If True, add offset of beta to normalized tensor. If False, beta is ignored.
batchnormalization_scale (boolean, default=True) – If True, multiply by gamma. If False, gamma is not used.
batchnormalization_beta_initializer (string/function, default='zeros') – Initializer for the beta weight.
batchnormalization_gamma_initializer (string/function, default='ones') – Initializer for the gamma weight.
batchnormalization_moving_mean_initializer (string/function, default='zeros') – Initializer for the moving mean.
batchnormalization_moving_variance_initializer (string/function,) – default=’ones’ Initializer for the moving variance.
batchnormalization_beta_constraint (function, default=None) – Optional constraint for the beta weight.
batchnormalization_gamma_constraint (function, default=None) – Optional constraint for the gamma weight.
dense_units (integer, default=None) – Dimensionality of the output space.
dense_activation (string/function, default='relu') – Activation function to use.
dense_use_bias (boolean, default=True) – Whether the layer uses a bias vector.
dense_kernel_initializer (string/function, default='he_uniform') – Initializer for the kernel weights matrix.
dense_bias_initializer (string/function, default='zeros') – Initializer for the bias vector.
dense_kernel_regularizer_l1 (float, default=None) – L1 regularization factor applied to the kernel weights matrix.
dense_kernel_regularizer_l2 (float, default=None) – L2 regularization factor applied to the kernel weights matrix.
dense_bias_regularizer_l1 (float, default=None) – L1 regularization factor applied to the bias vector.
dense_bias_regularizer_l2 (float, default=None) – L2 regularization factor applied to the bias vector.
dense_activity_regularizer_l1 (float, default=None) – L1 regularization factor applied to the output of the layer.
dense_activity_regularizer_l2 (float, default=None) – L2 regularization factor applied to the output of the layer.
dense_kernel_constraint (function, default=None) – Constraint function applied to the kernel weights matrix.
dense_bias_constraint (function, default=None) – Constraint function applied to the bias vector.
dropout_rate (float in [0, 1], default=0.0) – Fraction of the input units to drop.
dropout_noise_shape (array-like, default=None) – shape of the binary dropout mask that will be multiplied with the input.
dropout_seed (integer, default=None) – Random seed.
recurrent_regularizer_l1 (float, default=None) – L1 regularization factor applied to the recurrent_kernel weights matrix.
recurrent_regularizer_l2 (float, default=None) – L2 regularization factor applied to the recurrent_kernel weights matrix.
beta_regularizer_l1 (float, default=None) – L1 regularization factor applied to the beta weight.
beta_regularizer_l2 (float, default=None) – L2 regularization factor applied to the beta weight.
gamma_regularizer_l1 (float, default=None) – L1 regularization factor applied to the gamma weight.
gamma_regularizer_l2 (float, default=None) – L2 regularization factor applied to the gamma weight.
return_sequences (boolean, default=False) – Whether to return the last output in the output sequence, or the full sequence.