Machine Learning

A pragmatic definition...

Collection of algorithms and statistical models (methods) for machines to carry out automated tasks based on the observation of inputs and/or outputs of a process.

The goal of machine learning is to produce an estimation or a classification given a set of input values.

We often distinguish:

• ML method: the mechanism to train a model (neural network, support vector machine, etc).

• ML model: an instance of the method trained to replicate the behavior of the target process.

Types of Learning

Supervised

The model is trained with a dataset of the target process.

When trained for a classification task, the historical dataset should contain the Ground Truth - the actual class of a given sample.

Unsupervised

Classification or regression does not depend on prior knowledge.