Machine Learning in Science

AutoSBI Poster: Tuesday 2 Dec 10:30am at the Amortized ProbML workshop, Copenhagen 11/11

Fifth, we bring AutoML to SBI pipelines with a practical performance metric that does not require ground-truth posteriors, improving inference quality on the SBI benchmark! By @swagatam.bsky.social, @gmoss13.bsky.social, @keggensperger.bsky.social, @jakhmack.bsky.social 10/11

Kalman filtering meets Jaxley poster #2015: Thu 4 Dec 11:00am at San Diego ➡️ openreview.net/forum?id=1si... 9/11

Fourth, in collaboration with Ian C Tanoh and Scott Linderman, we used the Jaxley toolbox and extended Kalman filters to estimate the marginal log-likelihood of a biophysical neuron model. We showed that this enables identifying biophysical parameters given extracellular recordings. 8/11

Retina model with Jaxley poster #2015: Friday 5 Dec 11:00am at San Diego ➡️ openreview.net/forum?id=ayj... 7/11

Third, in collaboration with @kyrakadhim.bsky.social, @philipp.hertie.ai, and others, we built a task- and data-constrained biophysical network of the outer plexiform layer of the mouse retina. To optimize this model, we built it on top of our Jaxley toolbox for differentiable simulation. 6/11

NPE-PFN poster #509: Thursday 4 Dec 11:00am at San Diego and Thursday 4 Dec 10:30am at Copenhagen ➡️ openreview.net/forum?id=kN0... 5/11

Second, come by to check out NPE-PFN: We leverage the power of tabular foundation models for training-free and simulation-efficient SBI. SBI has never been so effortless! By @vetterj.bsky.social, Manuel Gloeckler, @danielged.bsky.social, @jakhmack.bsky.social 4/11

FNOPE poster #601: Friday 5 Dec 4:30pm at San Diego and Thursday 4 Dec 10:30am at Copenhagen ➡️ openreview.net/forum?id=yB5... 3/11

First, we introduce FNOPE, a new simulation-based inference approach for efficiently and flexibly inferring function-valued parameters. By @gmoss13.bsky.social, @leahsmuhle.bsky.social, Reinhard Drews, @jakhmack.bsky.social and @coschroeder.bsky.social 2/11

Our group is at NeurIPS and EurIPS this year with four papers and one workshop poster. If you are either curious about SBI with autoML, with foundation models, or on function spaces or about differentiable simulators with Jaxley, have a look below 👇 1/11

I’m super excited to present our new work in #Eurips2025 and #Neurips2025! We developed FNOPE: a new simulation-based inference (SBI) method which excels at inferring function-valued parameters!

Paper: openreview.net/forum?id=yB5...
Code: github.com/mackelab/fnope
(1/9)

We are looking for a Research Engineer (E13 TV-L) to work at the intersection of #ML and #compneuro! 🤖🧠

Help us build large-scale bio-inspired neural networks, write high-quality research code, and contribute to open-source tools like jaxley, sbi, and flyvis 🪰.

More info: www.mackelab.org/jobs/

Check out our website for the whole team: www.mackelab.org/people/ 7/7

Maren joined the lab as PhD student in October to work on connectome-constrained models of neural activity and behavior in the fruit fly. She holds an MSc Computational Neuroscience from the BCCN Berlin. 6/7

Byoungsoo (@byoungsookim.bsky.social) joined as a research scientist in October, upon finishing his Master's in Computational Neuroscience in the lab. He is working on modeling optomotor response circuits with a 3D compound eye model of the fruit fly. 5/7

Isaac joined as a master’s thesis student working on representation learning for connectome-constrained models. Now, as a PhD student since July, he’s applying this to models of the fruitfly. He previously did an MSc at AIMS South Africa. 4/7

Nicolas previously worked on computational neuroscience and NLP projects at EPFL. He joined the lab in June as a PhD student and is interested in building foundational models for neurophysiology data and applying LLMs for scientific discovery. 3/7

Stefan (@stewah.bsky.social) joined the lab as a PhD student in June. He completed his Bachelor’s and Master’s degrees in Physics at Heidelberg University. He works on using LLMs to discover scientific models. 2/7

MackeLab has grown! 🎉 Warm welcome to 5(!) brilliant and fun new PhD students / research scientists who joined our lab in the past year — we can’t wait to do great science and already have good times together! 🤖🧠 Meet them in the thread 👇 1/7

Nicolas has previously worked on computational neuroscience and NLP projects at EPFL. He joined the lab in June as a PhD student and is currently interested in building foundational models for neurological data and applying LLMs for scientific discovery. 3/7

Stefan (@stewah.bsky.social) joined the lab as a PhD student in June. He completed his Bachelor’s and Master’s degrees in Physics at Heidelberg University. At the moment, he is working on using LLMs to discover scientific models. 2/7

..., @ppjgoncalves.bsky.social, @janmatthis.bsky.social, @coschroeder.bsky.social, @jakhmack.bsky.social

Simulation-Based Inference: A Practical Guide A central challenge in many areas of science and engineering is to identify model parameters that are consistent with prior knowledge and empirical data. Bayesian inference offers a principled framewo...

Led by @deismic.bsky.social and @janboelts.bsky.social with Peter Steinbach, @gmoss13.bsky.social, @tommoral.bsky.social, Manuel Gloeckler, @plcrodrigues.bsky.social, Julia Linhart, @lappalainenjk.bsky.social, @bkmi.bsky.social,...

Finally, we built a database of SBI applications across fields as a community resource. We catalogued 100+ papers, available in an interactive web app. Users can explore what's been done, find similar problems to theirs, and contribute new applications: sbi-applications-explorer.streamlit.app

The appendix covers advanced topics and recent developments: sequential methods, model misspecification, function-valued parameters, score-based methods, flow matching. Extended practical guidelines help practitioners navigate the evolving SBI landscape.

We also cover a range of diagnostic tools that assess whether the posterior distributions returned by SBI are trustworthy. We discuss Posterior Predictive Checks and Calibration tests, as well as their advantages and limitations.

We provide three examples and code to demonstrate the complete workflow: gravitational wave parameter estimation (astrophysics), psychophysical model fitting (cognitive science), and ion channel inference (neuroscience).

We present a structured workflow with practical guidelines for each step: simulator setup, prior specification, method selection, network training, and validation. We illustrate each stage with concrete implementation details and common pitfalls.

This project brought together SBI researchers from 8 institutions across Europe—combining expertise from method developers and domain practitioners. Together, we aimed to consolidate years of experience into one comprehensive resource.