Dear students

welcome to the third lecture for the seminar

Advances in Deep Learning for Time Series.

Let me start again with an overview of the teaching team for this seminar.

My name is Dario Zank and I will give you the third lecture.

Zitin Jami from FAU, as well as me.

You will get to know later in this seminar.

And also Dr.

and Prof.

Christopher Loeffler from Pontificia Universidad Catholica de Valparaiso.

You can reach out to us and in the channels that we have built for communication.

You can ask questions whenever you need assistance.

Besides those occasions where we meet online or on campus to discuss topics about this seminar.

Today we are the third lecture and I will present to you deep learning for time series.

Meaning I will introduce the main architectures for deep neural networks that have been not necessarily designed for time series

but for sure used for modeling and processing data from the time series domain.

And hopefully provide examples for you to make it easy to understand when it's more convenient to use one or another.

References for these lectures are listed here.

So first of all the deep learning book from 2016 by a younger fellow, Joshua Benjamin R.

on Curveal.

It's called in fact deep learning and includes all the concepts that are at the basis of this discipline.

Including some description of the architectures that we will talk about today.

On a similar note the book by Bishop M.

Bishop, deep learning foundations and concepts.

This is a nice reading and part of the content of this lecture has also been taken or inspired by the content of this book.

As an outline of today's lecture

this is the content we will discuss today.

So first we will start with an introduction to deep learning.

I have to give a disclaimer that for this seminar that it's about advances in deep learning for time series

we assume that you all

you students have some knowledge about deep learning.

Because we never go in deep describing with enough time and enough depth all the concepts about deep learning

and especially all basic concepts about machine learning.

We have given some introduction to machine learning in the first lecture.

We will give some introduction to deep learning in this lecture.

But of course this is not all you need to know.

So if you are totally new to deep learning, this is also fine.

We try to make the lecture self inclusive

but take our notes as a starting point and try to dig deeper in concepts.

Because we are only scratching the surface.

So we go very quick when we introduce deep learning.

And you can also reach out to us if you want to have some more material to study those concepts more in depth.

After a brief introduction to deep learning

I will talk about the first deep learning architecture

convolutional neural networks.

Then in the third section

we will talk about recurrent models

meaning RNNs and LSTMs.

And finally

in the last section