Researcher profile

Sebastian Vogel

Sebastian Vogel contributes to research discovery and scholarly infrastructure.

ResearcherAffiliation not importedOpen to collaborate

Machine Learning cond-mat.mtrl-sci Neural and Evolutionary Computing

Trust snapshot

Quick read

Trust 15 - Baseline

3works

0followers

3topics

4close collaborators

Actions

Decide how to stay connected

Follow researcher0

Inspect adjacent work, topics, institutions and collaborators without jumping out to a separate graph page.

Building this graph slice

BZPEER is loading the nearby papers, people, topics and institutions for this page.

preprint2026arXiv

Kriging prior Regression: A Case for Kriging-Based Spatial Features with TabPFN in Soil Mapping

Machine learning and geostatistics are two fundamentally different frameworks for predicting and spatially mapping soil properties. Geostatistics leverages the spatial structure of soil properties, while machine learning captures the relationship between available environmental features and soil properties. We propose a hybrid framework that enriches ML with spatial context through engineering of 'spatial lag' features from ordinary kriging. We call this approach 'kriging prior regression' (KpR), as it follows the inverse logic of regression kriging. To evaluate this approach, we assessed both the point and probabilistic prediction performance of KpR, using the TabPFN model across six fieldscale datasets from LimeSoDa. These datasets included soil organic carbon, clay content, and pH, along with features derived from remote sensing and in-situ proximal soil sensing. KpR with TabPFN demonstrated reliable uncertainty estimates and more accurate predictions in comparison to several other spatial techniques (e.g., regression/residual kriging with TabPFN), as well as to established non-spatial machine learning algorithms (e.g., random forest). Most notably, it significantly improved the average R2 by around 30% compared to machine learning algorithms without spatial context. This improvement was due to the strong prediction performance of the TabPFN algorithm itself and the complementary spatial information provided by KpR features. TabPFN is particularly effective for prediction tasks with small sample sizes, common in precision agriculture, whereas KpR can compensate for weak relationships between sensing features and soil properties when proximal soil sensing data are limited. Hence, we conclude that KpR with TabPFN is a very robust and versatile modelling framework for digital soil mapping in precision agriculture.

preprint2019arXiv

High-pressure synthesis of ultraincompressible hard rhenium nitride pernitride Re$_{2}$(N$_{2}$)N$_{2}$ stable at ambient conditions

Here we report the synthesis of metallic, ultraincompressible (bulk modulus $K_{0}$ = 428(10) GPa) and very hard (nanoindentation hardness 36.7(8) GPa) rhenium (V) nitride pernitride Re$_{2}$(N$_{2}$)N$_{2}$. While the empirical chemical formula of the compound, ReN$_{2}$, is the same as for other known transition metals pernitrides, e.g. IrN$_{2}$, PtN$_{2}$, PdN$_{2}$ and OsN$_{2}$, its crystal chemistry is unique. The known pernitrides of transition metals consist of a metal in the oxidation state +IV and pernitride anions N$_{2}^{4-}$. ReN$_{2}$ contains both pernitride N$_{2}^{4-}$ and discrete N$^{3-}$ anions, which explains its exceptional properties. Moreover, in the original experimental synthesis of Re$_{2}$(N$_{2}$)N$_{2}$ performed in a laser-heated diamond anvil cell via a direct reaction between rhenium and nitrogen at pressures from 40 to 90 GPa we observed that the material was recoverable at ambient conditions. Consequently, we developed a route to scale up its synthesis through a reaction between rhenium and ammonium azide, NH$_{4}$N$_{3}$, in a large-volume press at 33 GPa. Our work resulted not only in a discovery of a novel material with unusual crystal chemistry and a set of properties attractive for potential applications, but also demonstrated a feasibility of surmounting conceptions common in material sciences.

preprint2016arXiv

Efficient Stochastic Inference of Bitwise Deep Neural Networks

Recently published methods enable training of bitwise neural networks which allow reduced representation of down to a single bit per weight. We present a method that exploits ensemble decisions based on multiple stochastically sampled network models to increase performance figures of bitwise neural networks in terms of classification accuracy at inference. Our experiments with the CIFAR-10 and GTSRB datasets show that the performance of such network ensembles surpasses the performance of the high-precision base model. With this technique we achieve 5.81% best classification error on CIFAR-10 test set using bitwise networks. Concerning inference on embedded systems we evaluate these bitwise networks using a hardware efficient stochastic rounding procedure. Our work contributes to efficient embedded bitwise neural networks.