In this seminar

we assume that you have a bit of knowledge about machine learning.

This is a nice to have preliminary knowledge

but we try to make it self-inclusive and therefore

these and the following sections will be about an introduction to machine learning.

We start from the general concept.

What is machine learning and what types of machine learning exist out there?

Under the big umbrella of AI that contains all the algorithms that mimic the intelligence of humans

able to resolve problems in ways that we consider smart from the simplest to the most complex

algorithm, under this big umbrella of AI we find machine learning.

Machine learning is

the set of methods that include algorithms that can parse data

that can learn from it

and then apply what they have learned to make informed decisions.

They use normally features

that in a more general way are extracted by humans or human experts

from the data and try to solve a problem in a data-driven way.

Within machine learning,

we identify even a subset of methods that is called deep learning.

Deep learning is based on the

neural networks algorithms.

They are inspired by the structure of

networks of neurons in our brain and they are a type of algorithm that can also learn from data

and solve tasks

but in this case features are not given

are not extracted normally by human experts

or are extracted in an automatic way.

We say that neural networks extract hidden features

and that discover hidden patterns and insights from the data and solve the task in a very fully

data-driven way.

So this is broadly the difference between AI, machine learning, and deep learning.

And now we move forward and talk about different types of learning algorithms.

So we are within

the machine learning umbrella and we distinguish between three main categories.

The three main

categories are supervised learning, unsupervised learning, and reinforcement learning.

Let's

address them one by one.

So in supervised learning, this is a type of learning that is

done using, exploiting the concept of a teacher.

We will see what that means.

It makes machine learning explicitly and it works with labeled data.

So in supervised learning

the agent observes some examples that are vectors in an input space or

feature vectors in an input space

but also the agent observes some output labels and this

input and output becomes in pairs and from this relationship

from the relationship between input

and output

the agent tries to learn and in particular