Zobrazeno 1 - 10
of 367
pro vyhledávání: '"Holzmüller, P."'
Autor:
Musekamp, Daniel, Kalimuthu, Marimuthu, Holzmüller, David, Takamoto, Makoto, Niepert, Mathias
Solving partial differential equations (PDEs) is a fundamental problem in engineering and science. While neural PDE solvers can be more efficient than established numerical solvers, they often require large amounts of training data that is costly to
Externí odkaz:
http://arxiv.org/abs/2408.01536
For classification and regression on tabular data, the dominance of gradient-boosted decision trees (GBDTs) has recently been challenged by often much slower deep learning methods with extensive hyperparameter tuning. We address this discrepancy by i
Externí odkaz:
http://arxiv.org/abs/2407.04491
Autor:
Zaverkin, Viktor, Holzmüller, David, Christiansen, Henrik, Errica, Federico, Alesiani, Francesco, Takamoto, Makoto, Niepert, Mathias, Kästner, Johannes
Publikováno v:
npj Comput. Mater. 10, 83 (2024)
Efficiently creating a concise but comprehensive data set for training machine-learned interatomic potentials (MLIPs) is an under-explored problem. Active learning, which uses biased or unbiased molecular dynamics (MD) to generate candidate pools, ai
Externí odkaz:
http://arxiv.org/abs/2312.01416
Publikováno v:
J. Chem. Phys. 156, 114103 (2022)
The accuracy of the training data limits the accuracy of bulk properties from machine-learned potentials. For example, hybrid functionals or wave-function-based quantum chemical methods are readily available for cluster data but effectively out-of-sc
Externí odkaz:
http://arxiv.org/abs/2312.01414
The success of over-parameterized neural networks trained to near-zero training error has caused great interest in the phenomenon of benign overfitting, where estimators are statistically consistent even though they interpolate noisy training data. W
Externí odkaz:
http://arxiv.org/abs/2305.14077
Autor:
Holzmüller, David, Bach, Francis
Sampling from Gibbs distributions $p(x) \propto \exp(-V(x)/\varepsilon)$ and computing their log-partition function are fundamental tasks in statistics, machine learning, and statistical physics. However, while efficient algorithms are known for conv
Externí odkaz:
http://arxiv.org/abs/2303.03237
Developing machine learning-based interatomic potentials from ab-initio electronic structure methods remains a challenging task for computational chemistry and materials science. This work studies the capability of transfer learning, in particular di
Externí odkaz:
http://arxiv.org/abs/2212.03916
Autor:
Viktor Zaverkin, David Holzmüller, Henrik Christiansen, Federico Errica, Francesco Alesiani, Makoto Takamoto, Mathias Niepert, Johannes Kästner
Publikováno v:
npj Computational Materials, Vol 10, Iss 1, Pp 1-18 (2024)
Abstract Efficiently creating a concise but comprehensive data set for training machine-learned interatomic potentials (MLIPs) is an under-explored problem. Active learning, which uses biased or unbiased molecular dynamics (MD) to generate candidate
Externí odkaz:
https://doaj.org/article/b80bd493f23e40339f5d153468a09724
Publikováno v:
Journal of Machine Learning Research, 24(164):1-81, 2023
The acquisition of labels for supervised learning can be expensive. To improve the sample efficiency of neural network regression, we study active learning methods that adaptively select batches of unlabeled data for labeling. We present a framework
Externí odkaz:
http://arxiv.org/abs/2203.09410
Artificial neural networks (NNs) are one of the most frequently used machine learning approaches to construct interatomic potentials and enable efficient large-scale atomistic simulations with almost ab initio accuracy. However, the simultaneous trai
Externí odkaz:
http://arxiv.org/abs/2109.09569