Sci → Soc: Building algorithms that can do something useful with data. My fascination for this field led me to my current PhD studies in machine learning at the University of Oxford at the Visual Geometry Group.
Sci → Sci: Mathematics is the language we use to systematically analyze nature and build models. I did a BSc in physics at the LMU Munich and a MSc in mathematical modelling at the University of Oxford.
Soc → Sci: To understand our economy and our society, we need better tools from network science and physics. I have been interested and working with (temporal) complex networks ever since my MSc.
In this paper we ask how many images are actually needed to learn the layers of a convolutional neural network using self-supervision. Our main result is that a few or even a single image together with strong data augmentation are sufficient to nearly saturate performance for early layers.
In this paper we investigate dietary transitions by analyzing a dataset of over 240 thousand recipes with 2.5M user ratings from a popular German recipe website with regards to sustainability trends.
Various time-series forecasting algorithms for 1-D data in python. In particular, the algorithms implemented/compared are: AR(I), Ridge Regression, Lasso Regression, RandomForestRegressor and LSTM.
In this project with the Potsdam Institute for Climate Impact Research, I extended a basic macroeconomic growth model to an agent-based model with social dynamics. This yields surprising results and at the same time alleviates some of the problems of normal representative agent models.