Features of Fitness Landscapes

Mentored by Ivo Sbalzarini & Peter F Stadler
at TU Dresden or Leipzig University

Fitness, energy, and cost landscapes are a central concept in many areas of science and computing. They are not only used to formalize loss functions in supervised learning and other complex optimization problems, but also frequently appear as models of biological evolution, molecular structure formation (protein and RNA folding), and chemical reaction networks.

Mathematically, fitness landscapes are functions over typically high-dimensional spaces equipped with some (semi-)norm and topological structure. Landscape features, defined jointly by the representation of the underlying entities, a notion of neighborhood or proximity, and the fitness function, determine the dynamics of folding, evolutionary optimization, and the performance of optimization algorithms. However, beyond global parameters such as landscape ruggedness, it remains poorly understood which features of a landscape exactly play a role, and how they can be estimated. In this project, we will exploit, modern techniques from AI and computational statistics to discover and characterize pertinent landscape features and develop methods that use knowledge of such features sampled at running time to guide and improve optimization algorithms.

Work Environment

You will have access to the machine-learning HPC resources of the ScaDS.AI Center for Scalable Data Analytics and AI Dresden/Leipzig, offering state-of-the-art CPU (Intel and IBM POWER 9) and GPU (Nvidia A100 and V100) resources, as well as high-performance computing and storage systems for computational experiments. You will also be working with the machine learning and AI people at the Center for Systems Biology Dresden (CSBD), a joint center between the Max Planck Society and TU Dresden, and the Interdisciplinary Center for Bioinformatics, similar joint center between the Max Planck Society and Leipzig University, where a workplace can be provided.

Prerequisites

The candidate will need a solid background in mathematics (an emphasis on high-dimensional statistics, functional analysis or topology will be an advantage). Since the work will require extensive computational experimentation, proficiency in scientific programming and interest in data analysis are necessary.