Previous work on learning physical systems from data has focused on
high-resolution grid-structured measurements. However, real-world knowledge of
such systems (e.g. weather data) relies on sparsely scattered measuring
stations. In this paper, we introduce a novel simulated benchmark dataset,
DynaBench, for learning dynamical systems directly from sparsely scattered data
without prior knowledge of the equations. The dataset focuses on predicting the
evolution of a dynamical system from low-resolution, unstructured measurements.
We simulate six different partial differential equations covering a variety of
physical systems commonly used in the literature and evaluate several machine
learning models, including traditional graph neural networks and point cloud
processing models, with the task of predicting the evolution of the system. The
proposed benchmark dataset is expected to advance the state of art as an
out-of-the-box easy-to-use tool for evaluating models in a setting where only
unstructured low-resolution observations are available. The benchmark is
available at https://anonymous.4open.science/r/code-2022-dynabench/.
Description
[2306.05805] DynaBench: A benchmark dataset for learning dynamical systems from low-resolution data
%0 Generic
%1 dulny2023dynabench
%A Dulny, Andrzej
%A Hotho, Andreas
%A Krause, Anna
%D 2023
%K benchmark deeplearning myown neuralpde
%T DynaBench: A benchmark dataset for learning dynamical systems from
low-resolution data
%U http://arxiv.org/abs/2306.05805
%X Previous work on learning physical systems from data has focused on
high-resolution grid-structured measurements. However, real-world knowledge of
such systems (e.g. weather data) relies on sparsely scattered measuring
stations. In this paper, we introduce a novel simulated benchmark dataset,
DynaBench, for learning dynamical systems directly from sparsely scattered data
without prior knowledge of the equations. The dataset focuses on predicting the
evolution of a dynamical system from low-resolution, unstructured measurements.
We simulate six different partial differential equations covering a variety of
physical systems commonly used in the literature and evaluate several machine
learning models, including traditional graph neural networks and point cloud
processing models, with the task of predicting the evolution of the system. The
proposed benchmark dataset is expected to advance the state of art as an
out-of-the-box easy-to-use tool for evaluating models in a setting where only
unstructured low-resolution observations are available. The benchmark is
available at https://anonymous.4open.science/r/code-2022-dynabench/.
@misc{dulny2023dynabench,
abstract = {Previous work on learning physical systems from data has focused on
high-resolution grid-structured measurements. However, real-world knowledge of
such systems (e.g. weather data) relies on sparsely scattered measuring
stations. In this paper, we introduce a novel simulated benchmark dataset,
DynaBench, for learning dynamical systems directly from sparsely scattered data
without prior knowledge of the equations. The dataset focuses on predicting the
evolution of a dynamical system from low-resolution, unstructured measurements.
We simulate six different partial differential equations covering a variety of
physical systems commonly used in the literature and evaluate several machine
learning models, including traditional graph neural networks and point cloud
processing models, with the task of predicting the evolution of the system. The
proposed benchmark dataset is expected to advance the state of art as an
out-of-the-box easy-to-use tool for evaluating models in a setting where only
unstructured low-resolution observations are available. The benchmark is
available at https://anonymous.4open.science/r/code-2022-dynabench/.},
added-at = {2023-09-19T11:00:35.000+0200},
author = {Dulny, Andrzej and Hotho, Andreas and Krause, Anna},
biburl = {https://www.bibsonomy.org/bibtex/2b9e82dea8ccee9c3f53dc52adb193b3a/annakrause},
description = {[2306.05805] DynaBench: A benchmark dataset for learning dynamical systems from low-resolution data},
interhash = {c005ae8a888da958aaede0341d45a099},
intrahash = {b9e82dea8ccee9c3f53dc52adb193b3a},
keywords = {benchmark deeplearning myown neuralpde},
note = {cite arxiv:2306.05805},
timestamp = {2023-09-19T11:00:35.000+0200},
title = {DynaBench: A benchmark dataset for learning dynamical systems from
low-resolution data},
url = {http://arxiv.org/abs/2306.05805},
year = 2023
}