Auxiliary Functions

Here are the auxiliary functions that are used in the rest of the programs. These functions can be divided in three groups:

  1. A function for dividing the data in batches.

  2. Some functions for evaluation predicted values returned by the network.

  3. Functions to load data (fashion and mnist).

deatf.auxiliary_functions.batch(data, n, i)

Extract from the given data a batch that go from the position i to the i + n; being n the batch size .

Parameters

  • data – Set of solutions intended to be fed to the network.

  • n – Size of the desired batch.

  • i – Index of the last solution used in the last epoch.

Returns

The batch of data form x with size n since the index i of the data.

deatf.auxiliary_functions.mse(true, prediction)

Calculates the mean squared error with numpy functions.

Parameters

  • true – True data, is the ground of truth in the evaluation.

  • prediction – Predicted data, is the data that is evaluated.

Returns

The mean squared error calculated between the true and predicted data.

deatf.auxiliary_functions.accuracy_error(true, prediction)

Calculates the accuracy error with numpy functions.

Parameters

  • true – True data, is the ground of truth in the evaluation.

  • prediction – Predicted data, is the data that is evaluated.

Returns

The accuracy error calculated between the true and predicted data.

deatf.auxiliary_functions.balanced_accuracy(true, prediction)

Calculates the balanced accuracy with numpy functions.

Parameters

  • true – True data, is the ground of truth in the evaluation.

  • prediction – Predicted data, is the data that is evaluated.

Returns

The balaced accuracy error calculated between the true and predicted data.

deatf.auxiliary_functions.load_fashion()

Loads and returns the data from the fashion mnist dataset and is returned already divided in train, validation and test.

Returns

Data of fashion mnist dataset divided in train, test and validation (x_train, y_train, x_test, y_test, x_val, y_val).

Rtype x_train

uint8 NumPy array of grayscale image data with shapes (42000, 28, 28), containing the training data.

Rtype y_train

uint8 NumPy array of labels (integers in range 0-9) with shape (42000,) for the training data.

Rtype x_test

uint8 NumPy array of grayscale image data with shapes (10000, 28, 28), containing the test data.

Rtype y_test

uint8 NumPy array of labels (integers in range 0-9) with shape (10000,) for the test data.

Rtype x_val

uint8 NumPy array of grayscale image data with shapes (18000, 28, 28), containing the validation data.

Rtype y_val

uint8 NumPy array of labels (integers in range 0-9) with shape (18000,) for the validation data.

deatf.auxiliary_functions.load_mnist()

Loads and returns the data from the mnist dataset and is returned already divided in train, validation and test.

Returns

Data of mnist dataset divided in train, test and validation (x_train, y_train, x_test, y_test, x_val, y_val).

Rtype x_train

uint8 NumPy array of grayscale image data with shapes (42000, 28, 28), containing the training data. Pixel values range from 0 to 255.

Rtype y_train

uint8 NumPy array of labels (integers in range 0-9) with shape (42000,) for the training data.

Rtype x_test

uint8 NumPy array of grayscale image data with shapes (10000, 28, 28), containing the test data. Pixel values range from 0 to 255.

Rtype y_test

uint8 NumPy array of labels (integers in range 0-9) with shape (10000,) for the test data.

Rtype x_val

uint8 NumPy array of grayscale image data with shapes (18000, 28, 28), containing the validation data. Pixel values range from 0 to 255.

Rtype y_val

uint8 NumPy array of labels (integers in range 0-9) with shape (18000,) for the validation data.