So good morning together and welcome to the second part of the Neural Networks lecture

or I have to say to the mathematical foundations of artificial intelligence, neural networks

and data analytics.

I'm keen on the formulation here.

Artificial intelligence is a generalization.

Neural networks is the mathematics in it and data analytics is the application of it.

The data might be numbers, which means sensor data or it might be images.

It might be acoustic data whatsoever.

But this is data.

Neural networks is the mathematics and this is clearly if I speak about foundations, mathematical

foundations, so my key point here is the neural network part.

But the guideline what we want to solve at the end of the day is the artificial intelligence

vision.

In the last lecture, we have focused on feedforward neural networks, which means we have focused

on systems which do function approximation.

You have an input, you have an output and then you study the interaction between inputs

and outputs.

Now in the second part here, we have recurrent neural networks, which means we try to study

dynamical systems, which from the mathematical viewpoint is a wonderful question and framework

to study.

And for the people who haven't seen me in the winter semester lecture here, so I'm chief

scientist for artificial intelligence at the Fraunhofer Institute in Nürnberg.

And here you have my email.

So to say something about myself and I believe that this is important to have this because

it's showing you my viewpoint on the subject.

I did study mathematics, computer science and economics at the University of Bonn, which

means my view on a subject, on forecasting, on dynamical systems and so on is the viewpoint

of mathematician.

But I always had in mind, you have to do something with this mathematics, for example, in economics

or wherever, computer science.

And from 1987 to 2017, I was in the Siemens corporate technology department in Munich.

I was one of the founders of the neural network research at Siemens.

And since 2000, I was senior research scientist for artificial intelligence at Siemens there.

And it's clear if you work for such a long time in an industry, on the one side, you

have to do good research because in the central research and development department of such

a large company, all the questions are finally ending, which are not solved in the business

department.

So it's clear that you are faced from morning to evening with complicated questions, which

means research is an important part in such a position.

But on the other side, you cannot do research at a green table.

You have to organize your research in the question that at the end of the day, something

is coming out which is reasonable for the company.

And so therefore, you have to have an application framework where you can say, yes, what I'm

doing here is not only good as a mathematical story, one can do something with it.

And my focus on this always was dynamical systems and forecasting.

And since 2017, I'm at Fraunhofer in Nuremberg.

From the content here, I'm doing practically the same, which means still artificial intelligence

neural networks and data analytics.

And if you have such a position, then you also have to have connections to the scientific

community.