Networks are used ubiquitously in community ecology, often to describe species-species

interactions. Such networks bridge the collection of data and the mathematical representation of

the ecosystem. This leads to an understanding of the diversity, dynamics and productivity of an

ecosystem. Such understanding is essential for a proper management and exploitation of

ecosystems. There is a great need for statistical and computational tools for such ecological

network data. In this project, I will investigate how machine learning methods can be used to

represent, understand, predict and control them. Firstly, I will explore how to construct a general

representation of these networks, which will provide the basis for analyzing these networks.

Furthermore, I will explore how these networks can be predicted based on traits of the individual

species and how this can be used to model network dynamics. Finally, the developed methodology

will be used to control and manage ecosystems. This project will provide ecologist and bioscience

engineers with valuable tools to answer both fundamental questions and to manage concrete

ecosystems. As I am focusing on tools for an abstract representation of these networks, the

methods are broadly relevant, for many different ecosystems, but also to other fields such as

social network analysis, biological network inference or collaborative filtering.