Deep learning techniques represent a certain part of wider machine learning methods and have become increasingly popular for a variety of real-life applications solving complex high-dimensional problems.

So far, typical applications for deep learning architectures such as deep neural networks and recurrent neural networks include speech and pattern recognition, language processing, audio recognition or machine translation. In all these applications, deep learning techniques were able to yield break-through results due to their highly flexible and innovative architectures and their approach of training models towards a set of given data.

Newly emerging generative techniques such as generative adversarial networks or variational autoencoders which had originally been developed for image generation purposes allow for powerful applications in the field of risk modelling and model validation.

The webinar will introduce these techniques in general and then present several use cases applying them to typical tasks from risk modelling and model validation. We will demonstrate where the sweet spot of these techniques is and how these techniques can be used to replace currently used methods and models or how they can be applied in combination with established statistical methods. Model validation builds another popular use case for such generative techniques where they are used as alternative benchmark to existing models with the beauty that the generative models introduced here are free of any strong assumptions and hence allow for the possibility of an “assumption-free model validation”.

The webinar will contain detailed case studies for demonstrating the capabilities of generative deep learning methods, such as applications to copulas, multivariate risk factor distributions, risk factor dimension reduction and detection of data errors.

This webinar is for practitioners in the area of risk modelling or model validation who want to gain a deeper understanding how cutting-edge deep learning techniques can be applied to this field.

Organised by the EAA - European Actuarial Academy GmbH.

The webinar is open to all interested persons.

Technical requirements and test session
Please check with your IT department if your firewall and computer settings support webinar participations (the programme GoToWebinar is used for the webinar). Please also make sure that you are joining the webinar with a stable internet connection. 

On 13 May 10:00 – 10:30 CEST, there will be a test session offered to all registered participants to test the software.

The webinar will introduce generative deep learning techniques in general and then present several use cases applying them to typical tasks from risk modelling and model validation.

We will demonstrate where the sweet spot of these techniques is and how these techniques can be used to validate or replace currently used methods and models or applied in combination with established statistical methods.

The webinar will contain detailed case studies for demonstrating the capabilities of generative deep learning methods, such as applications to copulas, multivariate risk factor distributions, risk factor dimension reduction and detection of data errors.

Dr Mario Hoerig, Partner, Oliver Wyman Actuarial
Mario Hoerig is a Partner with Oliver Wyman, co-leading the actuarial services offering in the German speaking markets. Mario focuses on quantitative modelling under Solvency II (economic scenario generators for risk-neutral and real-world purposes, ALM studies, risk factor modelling for Solvency II, risk aggregation, economic capital and capital management) and advises some of the largest insurance companies in Europe on these topics.

Dr Daniel Hohmann, Senior Manager, Oliver Wyman Actuarial
Daniel Hohmann is a Senior Manager with Oliver Wyman. He has a strong quantitative background and has been advising his clients on a variety of market risk and economic capital topics such as proxy modelling, risk-neutral and real-world scenario generation and time series analysis for market data.