A neural network, (also known as a Multi-layer Perceptron) is a mathematical model that is used in machine learning.

There are many ways to make a neural network learn, one of which is Supervised Learning.

The combination of the last layer’s Activation Function and Cost Function, determine the type of the problem you’re trying to solve, some examples include

• Binary Classification uses Sigmoid and Binary Cross-Entropy
  • The output is a number between 0 and 1, you can then classify both classes by checking if the value is $< 0.5$ or $> 0.5$
• Multi-label Classification uses Sigmoid and Binary Cross-Entropy
  • The output is a vector of probabilities, each indicating a different label

A neural network’s Forward-Propagation is used for predicting data and its Backward-Propagation used for training itself.

Neural Network Architecture