Learning deep
Benjamin Säfken, Alexander Silbersdorff, Christoph Weisser · 2020
The rise of artificial intelligence has come to the forefront of both academic and public discussion in recent years and accordingly the topic has gained considerable interest among students.Many of the recent advances in and the growing use of artificial intelligence are built around the applications of deep learning algorithms.In the winter semester 2018/19, we thus decided to offer a new seminar on deep learning algorithms to students of the masters programme in applied statistics at the University of Göttingen and did so again in the winter semester 2019/20.Following the Humboldtian model of higher education, we aimed to allow students to conduct their own research into the basic ideas, mechanics and practical applications of deep learning algorithms and thereby learn about the issue more deeply than by conventional lecture-based teaching.Their findings were presented in the seminar and subsequently portrayed in article-styled seminar papers.The results of this seminar, both in terms of the advancements made by many students in their understanding and the quality of many of the submitted seminar papers, were strikingly positive and deserving of publication in our eyes.Thus we decided to give the students the chance to publish their work in this edited volume.The seven best seminar papers were thus selected for publication and the selected students went through a full review process conducted by two researchers active in the field of deep learning.Upon successfully addressing the issues raised by the reviews, the articles were included in this volume.Given that the publication of the seminar papers was not our original intention, we left it up to the students to decide whether they would write the seminar paper in English or in German, with some groups choosing the former and some the latter option.Accordingly, the content of this book entails contributions in the two different languages.The contributions are structured as follows:The first paper by Clemens Haerder entitled "Deep Learning und Machine Learning: Ein Vergleich anhand des Boston Housing Value Datensatzes" provides an introductory contrast between simple neural nets and deep learning algorithms on the one III and depth of the understanding developed by students when left to explore deep learning algorithms in a deep manner following the Humboldtian model of higher education.We want to thank Dominik Becker who made considerable effort in formatting