I am a
machine learning researcher based in Copenhagen.
I work on algorithms for AI4Science, which means I do research at the interface of
numerical methods and machine learning,
including topics such as probabilistic numerics, differentiable programming, numerical linear algebra, differential equations, Bayesian machine learning, state-space models, Gaussian processes, or physics-informed machine learning
(Google Scholar).
I write a bunch of code, mainly in Python/JAX
(Github: @pnkraemer), for instance,
Matfree, Probdiffeq, and TUEplots.
If you need a three-sentence biography, e.g. for announcing a talk,
[click here].
I go by "Nico",
but use "Nicholas" when writing papers.
Online, I tend to be
@pnkraemer
(and "pn" stands for "Peter Nicholas", not for "probabilstic numerics").
Find my papers on Google Scholar and my code on GitHub.
Talk to me in English or German. I would love to hear from you.
November 2025:
Our preprint on "VIKING: Deep variational inference with stochastic projections" is on arXiv now!
Here is a link.
October 2025:
I'll be presenting a poster on the TMLR paper "Numerically Robust Fixed-Point Smoothing Without State Augmentation" at the ELLIS UnConference in Copenhagen.
Here is a link to the paper.
I'll also be attending EurIPS. Come and say hi!
October 2025:
Our preprint on "Matrix-Free Least Squares Solvers: Values, Gradients, and What to Do With Them" is on arXiv now!
Here is a link.
September 2025:
Our paper on approximate Bayesian neural operators has been accepted (and published) by TMLR. Here is a link to the paper, and here is a link to the bibtex.
September 2025:
I just returned from Nice, attending the International Conference on Probabilistic Numerics. There, I presented the paper on "Adaptive probabilistic ODE solvers without adaptive memory requirements" (here is a link). I also gave a demo of Probdiffeq, a JAX library for probabilistic solvers for differential equations. here is a link to the demo. Let me know what you think. Already looking forward to the next iteration of this event!
March 2025:
Our preprint on "Numerically robust Gaussian state estimation with singular observation noise" is on arXiv now!
Here is a link.
January 2025:
My paper on "Numerically robust fixed-point smoothing without state augmentation" has been now been published by TMLR!
Here
is a link to the PDF on OpenReview.
Here
is a link to the code on GitHub.
January 2025:
My paper on "Numerically robust fixed-point smoothing without state augmentation" has been accepted by TMLR! A link to the published paper will follow soon.
November 2024:
I will be again talking about adaptive ODE solvers, this time in Oxford.
November 2024:
I will be talking about adaptive ODE solvers at the Computer Lab in Cambridge. Here is a link to the announcement.
November 2024:
I will be visiting our CUQI-neighbours at DTU Compute to give a talk as a part of their seminar series.
October 2024:
Our paper on "Gradients of functions of large matrices" has been accepted as a spotlight at Neurips. Looking forward to presenting this in Vancouver!
Here is a link to the arXiv version.
October 2024:
My preprint on "Adaptive probabilistic ODE solvers without adaptive memory requirements" entered arXiv.
Here is a link.
October 2024:
My preprint on "Numerically robust fixed-point smoothing without state augmentation" entered arXiv.
Here is a link.
September 2024:
My preprint on "A tutorial on automatic differentiation with complex numbers" entered arXiv.
Here is a link.
September 2024:
If you're at GenU in Copenhagen (September 18-19), come and say hi! I'll bring a poster on computing gradients of functions of large matrices.
Here is a link to the preprint.
July 2024:
I am giving a talk on the "Probabilistic numerical method of lines", this time at
ProbNum24 in London
on July 15.
