Graph-partitioning based convolutional neural network  for earthquake detection using a seismic array

Yano, Keiji; Shiina, Takahiro; Kurata, Sumito; Kato, Aitaro; Komaki, Fumiyasu; Sakai, Shin’ichi; Hirata, Naoshi

doi:10.1002/essoar.10503223.1

Cited by 5 publications

(11 citation statements)

References 29 publications

(32 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The work of [5,14,15] comes most close to the goals of our work. Each of these papers tried to use time series related data in combination with graphs to improve predictions.…”

Section: Deep Learning For Seismic Analysismentioning

confidence: 65%

“…In other words, metadata was used to enhance their CNN model, except no graph layers were applied. [14] propose a technique that combines CNNs with graph partitioning to group time series together based on spatial information. This procedure increases the quality of the within-group features, and improves predictions, but no GNNs were utilised.…”

Section: Deep Learning For Seismic Analysismentioning

confidence: 99%

“…In contrast to previous research, in this paper we tackle the problem of time series regression with GNNs, for which we present an architecture to perform multivariate time series regression using GNNs. Previous attempts for tackling similar time series problems with graphbased methods have been made by [5,14,15], yet each have some shortcomings. [5] mention that they designed a Deep Learning approach using a GNN.…”

Section: Introductionmentioning

confidence: 99%

“…Therefore, while prediction scores improved, no actual GNN layers were used. Second, [14] proposed a graph partitioning algorithm that works together with a CNN. However, they make use of classical graph theory techniques and a GNN method is not applied.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Graph Neural Networks for Multivariate Time Series Regression with Application to Seismic Data

Bloemheuvel,

Hoogen,

Jozinović

et al. 2022

Preprint

View full text Add to dashboard Cite

Machine learning, with its advances in Deep Learning has shown great potential in analysing time series in the past. However, in many scenarios, additional information is available that can potentially improve predictions, by incorporating it into the learning methods. This is crucial for data that arises from e. g., sensor networks that contain information about sensor locations. Then, such spatial information can be exploited by modeling it via graph structures, along with the sequential (time) information. Recent advances in adapting Deep Learning to graphs have shown promising potential in various graph-related tasks. However, these

show abstract

“…The work of [5,14,15] comes most close to the goals of our work. Each of these papers tried to use time series related data in combination with graphs to improve predictions.…”

Section: Deep Learning For Seismic Analysismentioning

confidence: 65%

Section: Deep Learning For Seismic Analysismentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Graph Neural Networks for Multivariate Time Series Regression with Application to Seismic Data

Bloemheuvel,

Hoogen,

Jozinović

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…Previous attempts for tackling similar time series problems with graph-based methods have been made by [5,13,14], yet each has some shortcomings. van den Ende and Ampuero [5] mention that they designed a GNN for the localization of earthquakes from waveform data.…”

Section: Introductionmentioning

confidence: 99%

Graph neural networks for multivariate time series regression with application to seismic data

Bloemheuvel

Hoogen

Jozinović

et al. 2022

Int J Data Sci Anal

View full text Add to dashboard Cite

Machine learning, with its advances in deep learning has shown great potential in analyzing time series. In many scenarios, however, additional information that can potentially improve the predictions is available. This is crucial for data that arise from e. g., sensor networks that contain information about sensor locations. Then, such spatial information can be exploited by modeling it via graph structures, along with the sequential (time series) information. Recent advances in adapting deep learning to graphs have shown potential in various tasks. However, these methods have not been adapted for time series tasks to a great extent. Most attempts have essentially consolidated around time series forecasting with small sequence lengths. Generally, these architectures are not well suited for regression or classification tasks where the value to be predicted is not strictly depending on the most recent values, but rather on the whole length of the time series. We propose TISER-GCN, a novel graph neural network architecture for processing, in particular, these long time series in a multivariate regression task. Our proposed model is tested on two seismic datasets containing earthquake waveforms, where the goal is to predict maximum intensity measurements of ground shaking at each seismic station. Our findings demonstrate promising results of our approach—with an average MSE reduction of 16.3%—compared to the best performing baselines. In addition, our approach matches the baseline scores by needing only half the input size. The results are discussed in depth with an additional ablation study.

show abstract

MagInfoNet: Magnitude Estimation Using Seismic Information Augmentation and Graph Transformer

Chen

Wang

et al. 2022

Earth and Space Science

View full text Add to dashboard Cite

In this study, we propose a reliable data‐driven tool, MagInfoNet, to enhance the accuracy of magnitude estimation. Its architecture was assembled using the Pre‐Inform and Mag‐Pred modules to replace and update the key functions of traditional seismic analysis workflows. The Pre‐Inform module with the residual network was used for data pretreatment by combining the intrinsic characteristics of seismic signals with the potential features of the arrival and travel times. Meanwhile, using a graph transformer with an improved cyclic graph, the Mag‐Pred module was used to calculate magnitudes by the preprocessed information and the autocorrelation of seismic time series. Training and testing data were randomly selected from the Stanford Earthquake Data Set. The results show that the estimation accuracy, generalization, and robustness of the proposed MagInfoNet are better than those of three machine learning models. Besides, MagInfoNet can perform better for those samples with larger epicentral distances, enhancing the monitoring capacity of existing system for earthquake events in remote areas. Finally, we discuss the interpretability of the explainable MagInfoNet to verify the role of advanced neural network modules.

show abstract

Graph-partitioning based convolutional neural network for earthquake detection using a seismic array

Cited by 5 publications

References 29 publications

Graph Neural Networks for Multivariate Time Series Regression with Application to Seismic Data

Graph Neural Networks for Multivariate Time Series Regression with Application to Seismic Data

Graph neural networks for multivariate time series regression with application to seismic data

MagInfoNet: Magnitude Estimation Using Seismic Information Augmentation and Graph Transformer

Contact Info

Product

Resources

About