An artifical neural network is a computational model used in machine learning. By giving appropriate training data the neural network is able to learn and make predictions on new data it has not seen before.

In order to train the network stochastic gradient descent (SGD) is used. The equations used to implement SGD in the API are taken from the article 'Deriving the Backpropagation Equations from Scratch' by Thomas Kurbiel on www.towardsdatascience.com.

Activation functions currently supported in the API are:

• tanh
• relu
• sigmoid
• linear

Loss functions currently supported in the API are:

• MSE
• Binary Crossentropy