TALKS

Eine lange Zeit bildeten Recurrent Neural Networks die Grundlage für state-of-the-art Modelle zur Verarbeitung sequentieller Daten wie z.B. Texte oder Audiosignale. Für Aufgaben aus dem Bereich des Natural Language Processing wie maschinelle Übersetzungen, Sprachverständnis und Textgenerierung wurden diese kürzlich durch Modelle wie BERT, GPT-2 und XLNet abgelöst. Sie alle basieren auf der Transformer Architektur, die durch den Einsatz des Self-Attention-Mechanismus vollständig auf rekurrente und convolutional Layer verzichten kann.

Ziel des Vortrags ist, die Funktionsweise der Bausteine der Transformer Architektur anhand von Use-Case-Szenarien intuitiv verständlich zu machen, sowie deren Vor- und Nachteile im Vergleich zu RNNs zu erläutern.

Larger companies face the problem of large numbers of incoming documents, binding significant resources in reading and redirecting to the corresponding sections. In this presentation an approach is presented, which discusses the solution regarding incoming emails at ERGO Nuremberg. Approx. 1k emails receives this location per day, which were manually read and redirected. Expecting an estimated increase of the email rate, a solution based on current machine learning approaches has to be implemented to support the company. This talk will cover the data preparation, the architecture and the final operationalization of this model in detail. The additional challenges of data science in the field of insurances regarding data protection law will be discussed as well as the technical requirements and boundaries as well as the usual pitfalls.

Deutsche Telekom is the 4th biggest telecommunications company in the world, with millions of customers using its roaming mobile services every day. This presentation details how Deutsche Telekom builds and deploys machine learning capabilities for its commercial roaming business.
I will give an overview of our machine learning use-cases and dive into addressing the challenges of at-scale deployment including several aspects of business integration, data quality monitoring, feature engineering, machine learning model life cycle management, package management, process automation, code reproducibility and continuous integration. You will learn about the end-to-end machine learning workflow what we build to produce capabilities from forecasting to anomaly detection and integration of different open source data science / data visualization tools.
Finally I will cover lessons learned on challenges associated with enterprise IT integration of open source solutions, providing practical examples of pitfalls and how we addressed them.

A major requirement for Credit Scoring models is of course to provide a risk prediction that is as accurate as possible. In addition, regulators demand these models to be transparent and auditable. Thus, in Credit Scoring very simple Predictive Models such as Logistic Regression or Decision Trees are still widely used and the superior predictive power of modern Machine Learning algorithms cannot be fully leveraged. A lot of potential is therefore missed, leading to higher reserves or more credit defaults.

This talk presents an overview of techniques that are able to make “black box” machine learning models transparent and demonstrate how they can be applied in Credit Scoring. We use the DALEX set of tools to compare a traditional scoring approach with state of the art Machine Learning models and asses both approaches in terms of interpretability and predictive power. Results show that a comparable degree of interpretability can be achieved while machine learning techniques keep their ability to improve predictive power.