Postdoc in machine learning with Francis Bach & @GaelVaroquaux: neural networks, tabular data, uncertainty, active learning, atomistic ML, learning theory. https://dholzmueller.github.iohttps://bsky.app/profile/dholzmueller.bsky.socialhttps://bsky.app/profile/dholzmueller.bsky.social/post/3meo5xvpm7c25Super hyped that it's finally out! [contains quote post or other embedded content]12 Feb 2026 14:05 +0000at://did:plc:pf5dv2k4ev5wwfmxqilntimi/app.bsky.feed.post/3meo5xvpm7c25https://bsky.app/profile/dholzmueller.bsky.social/post/3lv3xtyqsik2gI got 3rd out of 691 in a tabular kaggle competition – with only neural networks! 🥉 My solution is short (48 LOC) and relatively general-purpose – I used skrub to preprocess string and date columns, and pytabkit to create an ensemble of RealMLP and TabM models. Link below👇29 Jul 2025 11:10 +0000at://did:plc:pf5dv2k4ev5wwfmxqilntimi/app.bsky.feed.post/3lv3xtyqsik2ghttps://bsky.app/profile/dholzmueller.bsky.social/post/3lupvpr3ct22rExcited to have co-contributed the SquashingScaler, which implements the robust numerical preprocessing from RealMLP! [contains quote post or other embedded content]24 Jul 2025 16:00 +0000at://did:plc:pf5dv2k4ev5wwfmxqilntimi/app.bsky.feed.post/3lupvpr3ct22rhttps://bsky.app/profile/dholzmueller.bsky.social/post/3lnjhsilt7c2z🚨ICLR poster in 1.5 hours, presented by @danielmusekamp.bsky.social : Can active learning help to generate better datasets for neural PDE solvers? We introduce a new benchmark to find out! Featuring 6 PDEs, 6 AL methods, 3 architectures and many ablations - transferability, speed, etc.!24 Apr 2025 00:38 +0000at://did:plc:pf5dv2k4ev5wwfmxqilntimi/app.bsky.feed.post/3lnjhsilt7c2zhttps://bsky.app/profile/dholzmueller.bsky.social/post/3ljzvx5a5222qPractitioners are often sceptical of academic tabular benchmarks, so I am elated to see that our RealMLP model outperformed boosted trees in two 2nd place Kaggle solutions, for a $10,000 forecasting challenge and a research competition on survival analysis. https://www.kaggle.com/competitions/rohlik-sales-forecasting-challenge-v2/discussion/56311710 Mar 2025 15:53 +0000at://did:plc:pf5dv2k4ev5wwfmxqilntimi/app.bsky.feed.post/3ljzvx5a5222qhttps://bsky.app/profile/dholzmueller.bsky.social/post/3ljko7aesf22dA new tabular classification benchmark provides another independent evaluation of our RealMLP. RealMLP is the best classical DL model, although some other recent baselines are missing. TabPFN is better on small datasets and boosted trees on larger datasets, though.04 Mar 2025 14:25 +0000at://did:plc:pf5dv2k4ev5wwfmxqilntimi/app.bsky.feed.post/3ljko7aesf22dhttps://bsky.app/profile/dholzmueller.bsky.social/post/3lfuaipnvt22xThe first independent evaluation of our RealMLP is here! On a recent 300-dataset benchmark with many baselines, RealMLP takes a shared first place overall. 🔥 Importantly, RealMLP is also relatively CPU-friendly, unlike other SOTA DL models (including TabPFNv2 and TabM). 🧵 1/16 Jan 2025 12:05 +0000at://did:plc:pf5dv2k4ev5wwfmxqilntimi/app.bsky.feed.post/3lfuaipnvt22xhttps://bsky.app/profile/dholzmueller.bsky.social/post/3ld4wzfvhos2uI'll present our paper in the afternoon poster session at 4:30pm - 7:30 pm in East Exhibit Hall A-C, poster 3304! [contains quote post or other embedded content]12 Dec 2024 18:54 +0000at://did:plc:pf5dv2k4ev5wwfmxqilntimi/app.bsky.feed.post/3ld4wzfvhos2u