Hauptinhalt

Topinformationen

Personen

Animal Communication meets Machine Learning

8.3201

Dozenten

Beschreibung

The past decades have seen an unprecedented increase in the recordings that researchers collect from communicating animals. As a result datasets are available for various species, including whales, dolphins, and marmosets, to name a few. Researchers that work with these data face multiple problems: First, the sheer amount of data is hard to handle in non-automated fashion. Second, the data need to be interpreted (and ideally be linked to behaviour). Third, when listening to the recordings, researchers use their human auditory system and therefore may miss important aspects not audible to them.

As a solution to these problem, machine learning techniques offer an interesting approach, as they can aid detection, classification, and clustering of animal vocalisations. In this practical course, a small group of ML-enthusiastic students will work side-by-side with researchers and build on an existing pipeline, previously developed for whale sounds by the research group of Prof. Kietzmann. We will explore deep neural networks for sound detection, as well as variational autoencoders for compressing and interpreting the sounds.

This course is a practical, which means that requirements include solid knowledge of python programming, machine learning techniques (with deep learning in particular) and maths (predominantly linear algebra). Interested students are asked to send a short (0.5 page) motivational letter that also describes how their previous experience or coursework fulfills the criteria listed above.

Weitere Angaben

Ort: 50/E04
Zeiten: Di. 14:00 - 16:00 (wöchentlich)
Erster Termin: Dienstag, 29.10.2024 14:00 - 16:00, Ort: 50/E04
Veranstaltungsart: Praktikum (Offizielle Lehrveranstaltungen)