# Seminar WS 17/18 Hauptseminar Numerik

## Neural Networks in Approximation Theory

Termin: | Donnerstags 14:15-15:00 (vorläufig) |

Vorbesprechung: | Donnerstag, 5.10.2017, 14:15 |

Ort: | Raum 6.020, Wegelerstr. 6 |

Anmeldung: | ab sofort per E-Mail an Jens Oettershagen |

### Background

We deal with the numerical approximation of multivariate functions, which often is a difficult problem because of the *curse of dimensionality*. In some cases, sparse grids and kernel methods can break this curse and allow for highly accurate approximations. Another approach that recently gained interest among computer scientists and statisticians are *deep neural networks* which can, at least in some applications, solve problems that were intractable so far.

In this seminar, we will discuss the mathematical foundations as well as some recent developments in the area of high-dimensional approximation.

Since the number of participants is limited, please register via email with Jens Oettershagen.