Overview

Module: ncxlib.neuralnetwork.neuralnet

Classes

NeuralNetwork(...) : Builds a Neural Network Class

Functions

Compile(...): Returns None. Compiles the Neural Network.

add_layer(...) : Adds a layer to the Neural Network

forward_propagate_all(...) : Forward Propagates all layers in layers list

forward_propagate_all_no_save(...) : Forward propagates all layers without saving layer information

back_propagation(...) : Performs the back propagation on all layers starting from the output layer

train(...) : Trains the model based on number of epochs, learning rate, input and labels

predict(...) : Predicts the output of the model

evaluate(...) : Provides an accuracy score across the model

save_model(...) : Saves the model as an .h5 into desired file path

load_model(...) : Loads the model back from the desired filepath. Must be an .h5

Module: ncxlib.neuralnetwork.activations

Classes

Activation(...) Abstract Base Class for all Activation classes

ReLU(...) : ReLU activation function class

LeakyReLU(...) : LeakyReLU activation function class

Sigmoid(...) : Sigmoid Activation function class

Softmax(...) : Softmax Activation function class

Tanh(...) : Tanh Activation function class

Module: ncxlib.neuralnetwork.layers

Classes

Layer(...) : Abstract Base Class for the Layer family

InputLayer(...) Input Layer Class to the Neural Network

FullyConnectedLayer(...) : FullyConnectedLayer (hidden layers) within the Neural Network

OutputLayer(...) : Output Layer of the Neural Network

Module: ncxlib.neuralnetwork.losses

Classes

LossFunction(...) : Abstract Base Class for all loss functions

MeanSquaredError(...) : Mean Squared Error loss function

BinaryCrossEntropy(...) : Binary Cross Entropy loss function

CategoricalCrossEntropy(...) : Categorical Cross Entropy loss function\

Module: ncxlib.neuralnetwork.optimizers

Classes

Optimizer(...) : Abstract Base Class for all optimizers

SGD(...) : Stochastic Gradient Descent Optimizer

SGDMomentum(...) : Stochastic Gradient Descent with Momentum Optimizer

RMSProp(...) : RMS Prop Optimizer

Adam(...) : Adam Optimizer

Module: ncxlib.neuralnetwork.initializers

Classes

Initializer(...) : Abstract Base Class for all Initializers

HeNormal(...) : He Normal Initializer

Zero(...) : Zero Initializer

Module: ncxlib.preprocessing

Classes

PreProcessor(...) : Abstract Base Class for all Preprocessors

OneHotEncoder(...) : One Hot Encoding Preprocessors

ImageRescaler(...) : Image Rescaler Preprocessor

ImageGrayscaler(...) : Image Grayscaler Preprocessor

Scaler(...) : Abstract Base Class for Scalers

MinMaxScaler(...) : Min-Max-Scaler Preprocessor

Module: ncxlib.dataloaders

Classes

DataLoader(...) : Abstract Base Class for all Dataloaders

CSVDataLoader(...) : CSV Data Loader

ImageDataLoader(...) : Image DataLoader

Module: ncxlib.generators

Functions

random_array(...) : Returns random uniform array of data

integer_array(...) : Returns random integer array

generate_training_data(...) : Generates random training data with num of samples, features, labels, etc.

Module: ncxlib.utils

Function

train_test_split(...) : splits the data into x_train, x_test, y_train, y_test

k_fold_cross_validation(...) : splits the data into k-folds for x_train, x_test, y_train, y_test and returns scores per fold