Workshop

Machine Learning is based on structures like Neural Networks, random forests, kernels, principal components and others. During a training phase parameters and hyperparameters of these structures are fitted in a best possible way to reflect the inherent dependencies hidden in the training data. This parameter fitting is usually done by solving an associated meta-optimization problem.

On the other hand, grey box approaches, which use knowledge based models and data simultaneously, are often the most suitable modelling class on which optimization algorithms might be based ready to compute good realizations of structures or processes in applications.

The course will address the question: »How to combine optimization and Machine Learning?« by discussing the structures of the underlying basic optimization problem arising in practice and the associated meta-optimization problem of statistical learning. We go two steps further by creating »art« with our obtained model, and by making optimal decisions based on our models. The latter are dynamical and stochastic, and appear in economics and finance.

In this workshop, three internationally recognized and leading scientists will introduce to state-of-the art theory and techniques as well as research challenges.