`mlair.model_modules.probability_models`¶

>>> MyCustomisedModel().model.compile(**kwargs) == MyCustomisedModel().compile(**kwargs)
True

Module Contents¶

Classes¶

`MyUnetProb`	The AbstractModelClass provides a unified skeleton for any model provided to the machine learning workflow.
`MyCNNProb`	Taken fromhttps://towardsdatascience.com/uncertainty-in-deep-learning-bayesian-cnn-tensorflow-probability-758d7482bef6
`VarDense`
`ProbTestModel`	The AbstractModelClass provides a unified skeleton for any model provided to the machine learning workflow.
`ProbTestModel2`	The AbstractModelClass provides a unified skeleton for any model provided to the machine learning workflow.
`ProbTestModel3`	The AbstractModelClass provides a unified skeleton for any model provided to the machine learning workflow.
`ProbTestModel4`	The AbstractModelClass provides a unified skeleton for any model provided to the machine learning workflow.
`ProbTestModelMixture`	The AbstractModelClass provides a unified skeleton for any model provided to the machine learning workflow.
`DenseVariationalCustom`	Trying to implement a DensVar that can be stored:
`Convolution2DReparameterizationCustom`

Functions¶

`prior_trainable`(kernel_size, bias_size=0, dtype=None)
`nll`(y_true, y_pred)	This function should return the negative log-likelihood of each sample
`posterior`(kernel_size, bias_size, dtype=None)
`prior`(kernel_size, bias_size, dtype=None)

Attributes¶

`__author__`
`__date__`
`tfd`
`tfb`
`tfpl`
`mylayer`

mlair.model_modules.probability_models.__author__ = Felix Kleinert¶

mlair.model_modules.probability_models.__date__ = 2022-07-08¶

mlair.model_modules.probability_models.tfd¶

mlair.model_modules.probability_models.tfb¶

mlair.model_modules.probability_models.tfpl¶

class mlair.model_modules.probability_models.MyUnetProb(input_shape: list, output_shape: list, num_of_training_samples: int)¶

Bases: mlair.model_modules.AbstractModelClass

The AbstractModelClass provides a unified skeleton for any model provided to the machine learning workflow.

The model can always be accessed by calling ModelClass.model or directly by an model method without parsing the model attribute name (e.g. ModelClass.model.compile -> ModelClass.compile). Beside the model, this class provides the corresponding loss function.

set_model(self)¶: Abstract method to set model.

set_compile_options(self)¶

This method only has to be defined in child class, when additional compile options should be used () (other options than optimizer and loss) Has to be set as dictionary: {‘optimizer’: None,

‘loss’: None, ‘metrics’: None, ‘loss_weights’: None, ‘sample_weight_mode’: None, ‘weighted_metrics’: None, ‘target_tensors’: None }

Returns

static prior(kernel_size, bias_size, dtype=None)¶

static posterior(kernel_size, bias_size, dtype=None)¶

class mlair.model_modules.probability_models.MyCNNProb(input_shape: list, output_shape: list)¶

Bases: mlair.model_modules.AbstractModelClass

Taken fromhttps://towardsdatascience.com/uncertainty-in-deep-learning-bayesian-cnn-tensorflow-probability-758d7482bef6 and modified to our data

static loss_fn(y_true, y_pred)¶

static custom_normal_prior(dtype, shape, name, trainable, add_variable_fn)¶

static laplace_prior(dtype, shape, name, trainable, add_variable_fn)¶

static approximate_kl(q, p, q_tensor)¶

conv_reparameterization_layer(self, filters, kernel_size, activation)¶

set_model(self)¶: Abstract method to set model.

set_compile_options(self)¶

This method only has to be defined in child class, when additional compile options should be used () (other options than optimizer and loss) Has to be set as dictionary: {‘optimizer’: None,

‘loss’: None, ‘metrics’: None, ‘loss_weights’: None, ‘sample_weight_mode’: None, ‘weighted_metrics’: None, ‘target_tensors’: None }

Returns

class mlair.model_modules.probability_models.VarDense(units, make_posterior_fn, make_prior_fn, kl_weight=None, kl_use_exact=False, activation=None, use_bias=True, activity_regularizer=None, **kwargs)¶

Bases: tensorflow.keras.layers.Layer

call(self, inputs)¶

get_config(self)¶

mlair.model_modules.probability_models.prior_trainable(kernel_size, bias_size=0, dtype=None)¶

class mlair.model_modules.probability_models.ProbTestModel(input_shape: list, output_shape: list)¶

Bases: mlair.model_modules.AbstractModelClass

The AbstractModelClass provides a unified skeleton for any model provided to the machine learning workflow.

The model can always be accessed by calling ModelClass.model or directly by an model method without parsing the model attribute name (e.g. ModelClass.model.compile -> ModelClass.compile). Beside the model, this class provides the corresponding loss function.

set_model(self)¶: Abstract method to set model.

set_compile_options(self)¶

This method only has to be defined in child class, when additional compile options should be used () (other options than optimizer and loss) Has to be set as dictionary: {‘optimizer’: None,

‘loss’: None, ‘metrics’: None, ‘loss_weights’: None, ‘sample_weight_mode’: None, ‘weighted_metrics’: None, ‘target_tensors’: None }

Returns

class mlair.model_modules.probability_models.ProbTestModel2(input_shape: list, output_shape: list)¶

Bases: mlair.model_modules.AbstractModelClass

The AbstractModelClass provides a unified skeleton for any model provided to the machine learning workflow.

The model can always be accessed by calling ModelClass.model or directly by an model method without parsing the model attribute name (e.g. ModelClass.model.compile -> ModelClass.compile). Beside the model, this class provides the corresponding loss function.

set_model(self)¶: Abstract method to set model.

set_compile_options(self)¶

This method only has to be defined in child class, when additional compile options should be used () (other options than optimizer and loss) Has to be set as dictionary: {‘optimizer’: None,

‘loss’: None, ‘metrics’: None, ‘loss_weights’: None, ‘sample_weight_mode’: None, ‘weighted_metrics’: None, ‘target_tensors’: None }

Returns

class mlair.model_modules.probability_models.ProbTestModel3(input_shape: list, output_shape: list)¶

Bases: mlair.model_modules.AbstractModelClass

The AbstractModelClass provides a unified skeleton for any model provided to the machine learning workflow.

The model can always be accessed by calling ModelClass.model or directly by an model method without parsing the model attribute name (e.g. ModelClass.model.compile -> ModelClass.compile). Beside the model, this class provides the corresponding loss function.

set_model(self)¶: Abstract method to set model.

set_compile_options(self)¶

This method only has to be defined in child class, when additional compile options should be used () (other options than optimizer and loss) Has to be set as dictionary: {‘optimizer’: None,

‘loss’: None, ‘metrics’: None, ‘loss_weights’: None, ‘sample_weight_mode’: None, ‘weighted_metrics’: None, ‘target_tensors’: None }

Returns

class mlair.model_modules.probability_models.ProbTestModel4(input_shape: list, output_shape: list)¶

Bases: mlair.model_modules.AbstractModelClass

The AbstractModelClass provides a unified skeleton for any model provided to the machine learning workflow.

The model can always be accessed by calling ModelClass.model or directly by an model method without parsing the model attribute name (e.g. ModelClass.model.compile -> ModelClass.compile). Beside the model, this class provides the corresponding loss function.

set_model(self)¶: Abstract method to set model.

set_compile_options(self)¶

This method only has to be defined in child class, when additional compile options should be used () (other options than optimizer and loss) Has to be set as dictionary: {‘optimizer’: None,

‘loss’: None, ‘metrics’: None, ‘loss_weights’: None, ‘sample_weight_mode’: None, ‘weighted_metrics’: None, ‘target_tensors’: None }

Returns

class mlair.model_modules.probability_models.ProbTestModelMixture(input_shape: list, output_shape: list)¶

Bases: mlair.model_modules.AbstractModelClass

The AbstractModelClass provides a unified skeleton for any model provided to the machine learning workflow.

The model can always be accessed by calling ModelClass.model or directly by an model method without parsing the model attribute name (e.g. ModelClass.model.compile -> ModelClass.compile). Beside the model, this class provides the corresponding loss function.

set_model(self)¶: Abstract method to set model.

set_compile_options(self)¶

This method only has to be defined in child class, when additional compile options should be used () (other options than optimizer and loss) Has to be set as dictionary: {‘optimizer’: None,

‘loss’: None, ‘metrics’: None, ‘loss_weights’: None, ‘sample_weight_mode’: None, ‘weighted_metrics’: None, ‘target_tensors’: None }

Returns

mlair.model_modules.probability_models.nll(y_true, y_pred)¶: This function should return the negative log-likelihood of each sample in y_true given the predicted distribution y_pred. If y_true is of shape [B, E] and y_pred has batch shape [B] and event_shape [E], the output should be a Tensor of shape [B].

mlair.model_modules.probability_models.posterior(kernel_size, bias_size, dtype=None)¶

mlair.model_modules.probability_models.prior(kernel_size, bias_size, dtype=None)¶

class mlair.model_modules.probability_models.DenseVariationalCustom¶

Bases: tfpl

Trying to implement a DensVar that can be stored: https://github.com/tensorflow/probability/commit/0ca065fb526b50ce38b68f7d5b803f02c78c8f16#

get_config(self)¶

class mlair.model_modules.probability_models.Convolution2DReparameterizationCustom¶

Bases: tfpl

get_config(self)¶

mlair.model_modules.probability_models.mylayer¶

mlair.model_modules.probability_models¶

Module Contents¶

Classes¶

Functions¶

Attributes¶

`mlair.model_modules.probability_models`¶