Model Module¶

class ModelAdapterBase¶

Bases: ABC

Implements a base class for model adapters. It provides an interface for loading, saving, printing summaries, fitting, predicting, and retrieving the model.

abstract fit(input_data: Any, output_data: Any, **kwargs) → Any¶

Fits the model to the provided input and output data.

Parameters:

input_data (Any) – The input data for fitting the model.
output_data (Any) – The output data for fitting the model.
(**kwargs) –
Additional keyword arguments for fitting the model.

Returns:

The result of the fitting process, which may vary depending on the model.

Return type:

(Any)

abstract get_model() → Any¶

Retrieves the underlying model.

Returns:: The underlying model instance.
Return type:: (Any)

abstract load_model(path: str) → None¶

Loads a model from the specified path.

Parameters:: path (str) – The path to the model file.

abstract predict(data: Any) → Any¶

Predicts the output for the given input data.

Parameters:: data (Any) – The input data for prediction.
Returns:: The predicted output data.
Return type:: (Any)

abstract print_summary(print_function: ~typing.Callable = <built-in function print>) → None¶

Prints a summary of the model’s architecture and parameters.

Parameters:: print_function (Callable) – The function to use for printing the summary. Defaults to the built-in print function.

report_parameters_needed_for_fitting() → list[str]¶

Reports the parameters needed for fitting the model.

Returns:: A list of parameter names that are required for fitting.
Return type:: (list[str])

abstract save_model(path: str) → None¶

Saves the model to the specified path.

Parameters:: path (str) – The path to the model file.

class SklearnModelAdapter(model: BaseEstimator)¶

Bases: ModelAdapterBase

Implements a model adapter for scikit-learn models. It provides methods for loading, saving, printing summaries, fitting, predicting, and retrieving the model.

TAG: str = 'sklearn'¶

fit(input_data: Any, output_data: Any, validation_data: Any, **kwargs) → dict¶

Fits the model to the provided input and output data.

Parameters:

input_data (Any) – The input data for fitting the model.
output_data (Any) – The output data for fitting the model.
validation_data (Any) – The validation data for evaluating the model.
(**kwargs) –
Additional keyword arguments for fitting the model.

Returns:

A dictionary containing the results of the fitting process.

Return type:

(dict)

get_model() → BaseEstimator¶

Retrieves the underlying model.

Returns:: The underlying model instance.
Return type:: (BaseEstimator)

load_model(path: str) → None¶

Loads a scikit-learn model from a file.

Parameters:

path (str) – The path to the model file.

Raises:

ValueError – If the path does not end with the expected model file extension.
FileNotFoundError – If the model file does not exist at the specified path.

predict(data: Any) → dict¶

Predicts probabilities for the output for the given input data.

Parameters:: data (Any) – The input data for prediction.
Returns:: The predicted output data.
Return type:: (dict)

print_summary(print_function: ~typing.Callable = <built-in function print>) → None¶

Prints a summary of the model’s name and parameters. Uses the provided print function to output the summary.

Parameters:: print_function (Callable) – The function to use for printing the summary.

report_parameters_needed_for_fitting() → list[str]¶

Reports the parameters needed for fitting the model.

Returns:: A list of parameter names that are required for fitting.
Return type:: (list[str])

save_model(path: str) → None¶

Saves a scikit-learn model to a file.

Parameters:: path (str) – The path to the model file.
Raises:: ValueError – If the path does not end with the expected model file extension.

class TFModelAdapter(model: Model, optimizer: Optional[OptimizerV2] = None, loss: Optional[str] = None, metrics: Optional[list[str]] = None)¶

Bases: ModelAdapterBase

Implements a model adapter for TensorFlow models. It provides methods for loading, saving, printing summaries, fitting, predicting, and retrieving the model.

TAG: str = 'tensorflow'¶

fit(input_data: Any, output_data: Any, validation_data: Any, epochs: int, batch_size: int, callbacks: list[tensorflow.python.keras.callbacks.Callback], **kwargs) → dict¶

Fits the model to the provided input and output data.

Parameters:

input_data (Any) – The input data for fitting the model.
output_data (Any) – The output data for fitting the model.
validation_data (Any) – The validation data for evaluating the model.
epochs (int) – The number of epochs to train the model.
batch_size (int) – The batch size to use for training.
callbacks (list[Callback]) – The list of callbacks to use during training.
(**kwargs) –
Additional keyword arguments for fitting the model.

Returns:

A dictionary containing the results of the fitting process.

Return type:

(dict)

get_model() → Model¶

Retrieves the underlying model.

Returns:: The underlying model instance.
Return type:: (Model)

load_model(path: str) → None¶

Loads a TensorFlow model from a file.

Parameters:: path (str) – The path to the model file.
Raises:: ValueError – If the path does not end with the expected weights file extension.

predict(data: Any) → dict¶

Predicts probabilities for the output for the given input data.

Parameters:: data (Any) – The input data for prediction.

print_summary(print_function: ~typing.Callable = <built-in function print>) → None¶

Prints a summary of the model’s architecture.

Parameters:: print_function (Callable) – The function to use for printing the summary.

report_parameters_needed_for_fitting() → list[str]¶

Reports the parameters needed for fitting the model.

Returns:: A list of parameter names that are required for fitting.
Return type:: (list[str])

save_model(path: str) → None¶

Saves a TensorFlow model to a file.

Parameters:: path (str) – The path to the model file.
Raises:: ValueError – If the path does not end with the expected weights file extension.

class GeneralSklearnBluePrint(base_estimator_class: type, **kwargs)¶

Bases: BluePrintBase

Implements a general blueprint for scikit-learn models. It provides a method to instantiate a model based on the provided base estimator class and keyword arguments.

instantiate_model(**kwargs) → ModelAdapterBase¶

Instantiates a model based on the provided keyword arguments.

Parameters:

(**kwargs) –

Optional keyword arguments for the model.

Returns:

The model adapter for the instantiated model.

Return type:

(ModelAdapterBase)

report_parameters_needed_for_instantiation() → list[str]¶

Reports the parameters needed for model instantiation.

Returns:: A list of parameter names that are required for instantiation.
Return type:: (list[str])

class VGGceptionCnnBluePrint(number_of_filters: int = 32, cnn_squeezing_coeff: int = 2, dense_squeezing_coeff: int = 2, dense_repetition_coeff: int = 1, filters_number_coeff: int = 2)¶

Bases: BluePrintBase

Blueprint for creating a hybrid CNN architecture combining VGG and Xception patterns.

This class implements a model blueprint that constructs a neural network with a combined architecture inspired by VGG16 and Xception networks. It’s designed to process both spatial and non-spatial features by separating the input vector and processing them through different network components before combining them.

instantiate_model(input_shape: tuple[int, int], output_length: int, spatial_data_shape: tuple[int, int], number_of_filters: Optional[int] = None, cnn_squeezing_coeff: Optional[int] = None, dense_squeezing_coeff: Optional[int] = None, dense_repetition_coeff: Optional[int] = None, filters_number_coeff: Optional[int] = None) → ModelAdapterBase¶

Creates and returns a hybrid VGG-Xception CNN model according to specified parameters.

The method constructs a neural network that: 1. Separates the input into spatial and non-spatial components 2. Processes the spatial data through VGG16 and Xception blocks 3. Flattens the CNN output and concatenates with non-spatial features 4. Passes the combined features through a series of dense layers 5. Produces a softmax output for classification

Parameters:

input_shape (tuple[int, int]) – Shape of the input tensor
output_length (int) – Number of output classes/actions
spatial_data_shape (tuple[int, int]) – Rows and columns to reshape spatial data
number_of_filters (int) – Initial number of convolutional filters
cnn_squeezing_coeff (int) – Factor by which CNN dimensions are reduced
dense_squeezing_coeff (int) – Factor by which dense layer sizes are reduced
dense_repetition_coeff (int) – Number of dense layers of the same size to use
filters_number_coeff (int) – Factor by which filter count increases in convolutional layers

Returns:

Keras model implementing the hybrid VGG-Xception architecture to be compiled further.

Return type:

Model

report_parameters_needed_for_instantiation() → list[str]¶

Reports the parameters needed for model instantiation.

Returns:: A list of parameter names that are required for instantiation.
Return type:: (list[str])

class InceptionBlock(kernels: tuple[tuple[int, int], tuple[int, int], tuple[int, int], tuple[int, int]], filters: tuple[int, int, int, int], steps: tuple[int, int])¶

Bases: object

Class implementing an Inception block compatible with the TensorFlow API. This block uses parallel convolutions with different kernel sizes and a max-pooling layer, followed by a concatenation of the results, as seen in the Inception architecture.

Diagram:

Input Tensor --> +-----------------------+
    |            | Conv2D - typ. 1xY     |                               +-------------+
    |            | Kernel Size: K1xK1    |------------------------------>| Concatenate |
    |            | Filters: N1           |                               |             |
    |            +-----------------------+                               |             |
    |                                                                    |             |
    +----------> +-----------------------+   +-----------------------+   |             |
    |            | Conv2D - typ. 3xY     |   | Conv2D - typ. 3xY     |   |             |
    |            | Kernel Size: K1xK1    |-->| Kernel Size: K2xK2    |-->|             |
    |            | Filters: N2           |   | Filters: N2           |   |             |
    |            +-----------------------+   +-----------------------+   |             |
    |                                                                    |             |
    +----------> +-----------------------+   +-----------------------+   |             |
    |            | Conv2D - typ. 5xY     |   | Conv2D - typ. 5xY     |   |             |
    |            | Kernel Size: K1xK1    |-->| Kernel Size: K3xK3    |-->|             |
    |            | Filters: N3           |   | Filters: N3           |   |             |
    |            +-----------------------+   +-----------------------+   |             |
    |                                                                    |             |
    +----------> +-----------------------+   +-----------------------+   |             |
                 | MaxPooling2D          |   | Conv2D                |   |             |
                 | Kernel Size: K4xK4    |-->| Kernel Size: K1xK1    |-->|             |
                 | Stride: S1xS1         |   | Filters: N4           |   +-------------+ --> Output Tensor
                 +-----------------------+   +-----------------------+

class SEBlock(reduction_ratio: int = 16)¶

Bases: object

Class implementing a Squeeze-and-Excitation (SE) block compatible with the TensorFlow API. This block applies global average pooling followed by a squeeze-and-excitation operation, as described in the SE-Net architecture.

Diagram:

Input Tensor last dimension length - ITldl
Reduction rate - Rr

Input Tensor --> +-----------------+   +---------------+   +--------------------+   +--------------+
    |            | GlobalAvgPool   |-->| Reshape       |   | Dense              |   | Dense        |   +----------+
    |            +-----------------+   | Shape: ITldl  |-->| Nodes: ITldl // Rr |-->| Nodes: ITldl |-->| Multiply |
    |                                  |               |   |                    |   |              |   |          |
    |                                  +---------------+   +--------------------+   +--------------+   |          |
    |                                                                                                  |          |
    +------------------------------------------------------------------------------------------------->|          |
                                                                                                       +----------+ --> Output Tensor

class Vgg16Block(kernels: tuple[tuple[int, int], tuple[int, int], tuple[int, int]], filters: tuple[int, int])¶

Bases: object

Class implementing Vgg16 block compatible with tensorflow API. This block is a core component of the VGG16 architecture, applying two convolutional layers followed by a max pooling layer to downsample and extract features from the input tensor.

Diagram:

Input Tensor --> +-----------------------+   +-----------------------+   +-----------------------+
                 | Conv2D                |   | Conv2D                |   | MaxPooling2D          |
                 | Filters: N1           |-->| Filters: N2           |-->| Pool Size: K3xK3      |
                 | Kernel Size: K1xK1    |   | Kernel Size: K2xK2    |   |                       |
                 +-----------------------+   +-----------------------+   +-----------------------+ --> Output Tensor

class XceptionBlock(kernels: tuple[tuple[int, int], tuple[int, int], tuple[int, int], tuple[int, int]], filters: tuple[int, int, int], steps: tuple[tuple[int, int], tuple[int, int]])¶

Bases: object

Class implementing an Xception block compatible with the TensorFlow API. This block implements depthwise separable convolutions followed by max pooling and a residual connection, as seen in the Xception architecture.

Diagram:

Input Tensor --> +-----------------------+   +----------------------+   +--------------------+   +-----+
    |            | SeparableConv2D       |   | SeparableConv2D      |   | MaxPooling2D       |   | Add |
    |            | Filters: N1           |-->| Filters: N2          |-->| Pool Size: K3xK3   |-->|     |
    |            | Kernel Size: K1xK1    |   | Kernel Size: K2xK2   |   | Stride: S1xS1      |   |     |
    |            +-----------------------+   +----------------------+   +--------------------+   |     |
    |                                                                                            |     |
    +----------> +-----------------------+                                                       |     |
                 | Conv2D                |                                                       |     |
                 | Filters: N3           |                                                       |     |
                 | Kernel Size: K4xK4    |------------------------------------------------------>|     |
                 | Stride: S2xS2         |                                                       |     |
                 |                       |                                                       +-----+ --> Output Tensor
                 +-----------------------+

Crypto Evolution

Navigation

Related Topics

Model Module¶