Okay, hello, welcome everyone, good evening.

So this will be the lecture Machine Learning for Physicists.

My name is Florian Markwardt, I'm here at the Max Planck Institute for the Science of Light in Erlang and also at the University.

So, regarding today, I will first give the lecture, the first lecture of this course,

and then we will have some time also to discuss maybe some of the more organizational things.

Please note that the lecture will be recorded. I mean, you are not in the picture, just I am in the picture.

Also, there's a website which I will direct you to afterwards,

and there will be tutorials for this lecture offered as well for you,

and they will be hands-on.

So the whole principle of this lecture is to give you a hands-on experience with neural networks,

so that in the end you can really program even quite advanced stuff.

Okay.

So let us start.

If you think of your brain,

it is quite an amazing piece of software or hardware as the case may be.

And it is quite amazing in many respects,

but especially also because it is very flexible.

And if I am a physicist or mathematician

and I want to boil down the workings of the brain to the most elementary model,

I could say the brain is something which takes an input and generates an output.

The output might be, for example, what I will talk to you, or how I move around in the world.

Now, how does that work inside your brain?

And the amazing thing is that it works by electricity, by electrical signals.

And this was realized already hundreds of years ago.

But, the microscopic details started to be understood about a hundred years ago.

And he has an actual drawing by semin fragments Spain.

He was one of the pioneers of investigating neurons in the brain and so

here you see that these are the bodies of these neuro Ashadus and then there

are many branches of which they can be connected to other neurons and then

that will be electrical signals traveling more these neurons and you can

you these norms are kinds of which is and so if one gets activated maybe

maybe another neuron also gets activated and so on.

So this lecture will not be about the biological workings of neurons.

But of course,

this has inspired a lot of mathematicians and

computer scientists,

inspired them to try to invent something that is artificial,

that works as an algorithm on a real computer as we know it,

but that can be as flexible as the human mind because in particular,

what we want is something that works

without you having to prescribe a detailed algorithm.

But instead it should be able to learn on its own,

just as your brain is able to learn on its own.

So that's the purpose of inventing

what is called artificial neural networks.

So they are inspired by how our brain works.

They do not exactly model how our brain works,

but their goal is also to be as flexible

and to be able to figure out new stuff

by looking at training examples essentially.