←
Jonas Dippel57 posts
Jonas Dippel avatar

Jonas Dippel

@jdppel

PhD Student at @ml_tuberlin, @bifoldberlin and Data Scientist at @aignostics | Training large neural nets for computational pathology.

◷ Joined December 2014▧ 1 Photos and videos
Laure Ciernik avatar
⟳ Jonas Dippel retweeted
Laure Ciernik@lciernik·6 Jun 2025

🎉 Update: This work got accepted to #icml2025!! Huge thanks to my amazing co-authors @LorenzLinhardt, Marco Morik, @jdppel, @skornblith, and @lukas_mut for their great work and to all collaborators! 🙏 📄 Paper: arxiv.org/abs/2411.05561 💻 Code: github.com/lciernik/similari… 🧵1/3

Post attachment

If two models are more similar to each other than a third on ImageNet, will this hold for medical/ satellite images? Our preprint analyzes how vision model similarities generalize across datasets, the factors that influence them, and their link to downstream task behavior. 🧵1/7

Mayo Clinic avatar
⟳ Jonas Dippel retweeted
Mayo Clinic@MayoClinic·13 Jan 2025

Mayo Clinic announced the formation of Mayo Clinic Digital Pathology, designed on a platform architecture to boldly unlock the power of its extensive archive of digital slides to revolutionize pathology and accelerate medical breakthroughs. Learn more: mayocl.in/4aeawwi

Post attachment
NEJM AI avatar
⟳ Jonas Dippel retweeted
NEJM AI@NEJM_AI·12 Nov 2024

A deep anomaly detection approach for histopathology shows high detection performance for a broad range of diseases (including all cancers) within the long diagnostic tail in gastrointestinal biopsies. Read the full article by @jdppel et al.: nejm.ai/3YwvUZz

Post attachment
Laure Ciernik avatar
⟳ Jonas Dippel retweeted
Laure Ciernik@lciernik·11 Nov 2024

If two models are more similar to each other than a third on ImageNet, will this hold for medical/ satellite images? Our preprint analyzes how vision model similarities generalize across datasets, the factors that influence them, and their link to downstream task behavior. 🧵1/7

Jonas Dippel avatar
Jonas Dippel@jdppel·25 Oct 2024

🙏 Thanks to my co-first author @n_prenissl , stellar supervision by @lukasruff , Klaus-Robert Müller, @FKlauschen and amazing contributions by @hense96, Philipp Liznerski, Tobias Winterhoff, Simon Schallenberg, Marius Kloft, Oliver Buchstab, David Horst, and Maximilian Alber.

Jonas Dippel avatar
Jonas Dippel@jdppel·25 Oct 2024

✨On two large clinical cohorts of gastrointestinal biopsies, we demonstrate that our anomaly detection approach can reliably detect a broad spectrum of anomalous pathologies, including rare cancer types.

Jonas Dippel avatar
Jonas Dippel@jdppel·25 Oct 2024

🛠️We compare two promising approaches. (1) using representation distances from SSL/foundation models and (2) contrasting normal examples with images from other tissue types which helps to learn a tight decision boundary around the normal data (outlier exposure).

Jonas Dippel avatar
Jonas Dippel@jdppel·25 Oct 2024

💡Detecting rare diseases is challenging as often not enough training data is available to train a supervised model. We circumvent this by proposing an anomaly detection approach which learns the characteristics of healthy tissue and common findings and then detects any deviation

Frieda Born avatar
⟳ Jonas Dippel retweeted
Frieda Born@FriedaBorn·15 Oct 2024

Why do some memories fade in seconds, while others stay with us for life? Working Memory (WM) holds info for just moments, but certain bits manage to stick around and make it into Long-Term Memory (LTM). In our new ⚡️preprint⚡️, we examined what helps these memories stick. 1/

Post attachment
Lukas Muttenthaler avatar
⟳ Jonas Dippel retweeted
Lukas Muttenthaler@lukas_mut·28 Sep 2023

The camera-ready version is now available on arXiv! 🔥 We've updated our manuscript with results for a few interesting experiments that were suggested by the reviewers during the rebuttal period. See the thread below👇

Happy to announce that this wonderful collaborative effort was accepted to @NeurIPSConf! Congrats team! Stay tuned for cam ready 🔥🧠🤖