Kinetic turbulence in magnetized space plasmas has been extensively studied via in situ observations, numerical simulations, and theoretical models. In this context, a key point concerns the formation of coherent current structures and their disruption through magnetic reconnection. We present automatic techniques aimed at detecting reconnection events in a large data set of numerical simulations. We make use of clustering techniques known as K-means and DBscan (usually referred to in literature as unsupervised machine-learning approaches), and other methods based on thresholds of standard reconnection proxies. All our techniques also use a threshold on the aspect ratio of the regions selected. We test the performance of our algorithms. We propose an optimal aspect ratio to be used in the automated machine-learning algorithm: AR = 18. The performance of the unsupervised approach results in it being strongly competitive with respect to those of other methods based on thresholds of standard reconnection proxies
Detecting Reconnection Events in Kinetic Vlasov Hybrid Simulations Using Clustering Techniques
Sisti, Manuela;Finelli, Francesco;Califano, Francesco;
2021-01-01
Abstract
Kinetic turbulence in magnetized space plasmas has been extensively studied via in situ observations, numerical simulations, and theoretical models. In this context, a key point concerns the formation of coherent current structures and their disruption through magnetic reconnection. We present automatic techniques aimed at detecting reconnection events in a large data set of numerical simulations. We make use of clustering techniques known as K-means and DBscan (usually referred to in literature as unsupervised machine-learning approaches), and other methods based on thresholds of standard reconnection proxies. All our techniques also use a threshold on the aspect ratio of the regions selected. We test the performance of our algorithms. We propose an optimal aspect ratio to be used in the automated machine-learning algorithm: AR = 18. The performance of the unsupervised approach results in it being strongly competitive with respect to those of other methods based on thresholds of standard reconnection proxiesFile | Dimensione | Formato | |
---|---|---|---|
Sisti ApJ2021.pdf
solo utenti autorizzati
Descrizione: Versione Editoriale finale
Tipologia:
Versione finale editoriale
Licenza:
NON PUBBLICO - Accesso privato/ristretto
Dimensione
1.75 MB
Formato
Adobe PDF
|
1.75 MB | Adobe PDF | Visualizza/Apri Richiedi una copia |
2012.09683.pdf
Open Access dal 18/02/2022
Tipologia:
Documento in Post-print
Licenza:
Creative commons
Dimensione
1.71 MB
Formato
Adobe PDF
|
1.71 MB | Adobe PDF | Visualizza/Apri |
Sisti_ApJ_unsupervised_ML_referee_resub.pdf
accesso aperto
Tipologia:
Documento in Pre-print
Licenza:
Creative commons
Dimensione
1.67 MB
Formato
Adobe PDF
|
1.67 MB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.