Glossary

In Supervised Learning, the training dataset includes inputs and also outputs, that the model uses to compare to its predicted outputs.

A node or neuron is a unit of computation of a Neural Network. It includes a set of weights associated to its inputs, and an Activation Function.

A row of nodes is a layer.

A Neural Network is a set of interconnected nodes, designed to recognize patterns from numerical data.

If it has layers between the input and output it's called a Deep Neural Network.

A row of nodes in between the input and output is a Hidden Layer.

The Activation Function is the function in each node that's responsible for its activation. It takes the sum of the weights as inputs and outputs a signal.

The Loss is the measure of how far away the predicted outputs are from the true outputs. It is calculated by a Loss Function.

The Optimizer is the algorithm used to minimize the Loss. It changes the weights of the nodes.

A batch is a subset of our dataset. At the end of each batch, the Optimizer updates the networks weights.

Every time the Optimizer works through the dataset, it's called an Epoch. Every epoch is subdivided in batches.

A Neural Network where the information travels only from input to output, with no loops, is Sequential (or Feedforward).

One-hot Encoding is a type of data encoding where categorical variables are converted into several variables with binary values.

See Classifying Iris Flowers for an example.