Nikolas Kirschstein

Hi, I hope you’re having a wonderful day! I’m a research-focused maths and informatics student enthusiastic about statistical machine learning, reinforcement learning, causal inference, and randomised algorithms. And I make a point of pointing out that calling informatics “computer science” would be like calling astrophysics “telescope science”… 😉

One of my favourite formulae is the bias-variance decomposition in least-squares regression with risk \(r(a) = \mathbb{E}[(a(X) - Y)^2]\), computed (randomised) algorithm \(A\), and optimal algorithm \(a^*\):

\[\underbrace{\mathbb{E}[r(A)] - r(a^*)}_{\text{expected excess risk}} = \underbrace{\mathbb{E}\left[(\mathbb{E}[A(X) \mid X] - a^*(X))^2\right]}_{\text{expected square bias}} + \underbrace{\mathbb{E}[\mathrm{Var}[A(X) \mid X]]}_{\text{expected variance}}\]

Education

MSc in Informatics, TUM, 2025 (expected)
MSc in Mathematics, Oxford University, 2023
BSc in Mathematics, University of Passau, 2022
BSc in Informatics, University of Passau, 2022

Work experience

Summer 2023: Machine Learning Engineer Intern, Google, California & New York
Summer 2022: Software Engineering Intern, Google, San Francisco
2018-2022: Self-employed Web Developer, Remote, Germany

Publications

The Merit of River Network Topology for Neural Flood Forecasting

Kirschstein, N., Sun, Y. (2024). The Merit of River Network Topology for Neural Flood Forecasting. ICML 2024.

Deep Active Learning for Detection of Mercury’s Bow Shock and Magnetopause Crossings

Julka, S., Kirschstein, N., Granitzer, M., Lavrukhin, A., Amerstorfer, U. (2023). Deep Active Learning for Detection of Mercury’s Bow Shock and Magnetopause Crossings. In: Amini, MR., Canu, S., Fischer, A., Guns, T., Kralj Novak, P., Tsoumakas, G. (eds) Machine Learning and Knowledge Discovery in Databases. ECML PKDD 2022. Lecture Notes in Computer Science(), vol 13716. Springer, Cham.

Deep Active Learning with Artificial Neural Networks for Automatic Detection of Mercury’s Bow Shock and Magnetopause Crossing Signatures in NASA’s MESSENGER Magnetometer Observations