Multitask Learning for Estimation of Magnetic Parameters Using Pattern Recognition
Machine learning (ML) approaches present an effective technique for accurately and efficiently predicting device parameters. Using these techniques, we introduce a multi-task convolutional neural network (CNN) model and support vector regression (SVR) model that is intended to precisely estimate two...
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2024-01-01
|
| Series: | IEEE Open Journal of Nanotechnology |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10748362/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1832590326813949952 |
|---|---|
| author | Anubha Sehgal Shipra Saini Hemkant Nehete Kunal Kranti Das Sourajeet Roy Brajesh Kumar Kaushik |
| author_facet | Anubha Sehgal Shipra Saini Hemkant Nehete Kunal Kranti Das Sourajeet Roy Brajesh Kumar Kaushik |
| author_sort | Anubha Sehgal |
| collection | DOAJ |
| description | Machine learning (ML) approaches present an effective technique for accurately and efficiently predicting device parameters. Using these techniques, we introduce a multi-task convolutional neural network (CNN) model and support vector regression (SVR) model that is intended to precisely estimate two important parameters of magnetic systems such as the Dzyaloshinskii-Moriya interaction (DMI) constant and the exchange constant (A<sub>ex</sub>). The magnetic Hamiltonian encapsulates various energy components, including exchange energy, DMI, Zeeman energy, and anisotropy energy, wherein factors such as saturation magnetization, DMI strength, exchange stiffness, and anisotropy constants influence their magnitudes. Conventionally, the estimation of these parameters has been computationally intensive and time-consuming. The CNN and SVR models can simultaneously estimate both the DMI constant and the exchange constant, making it a versatile tool for magnetic system characterization. The custom CNN model performs best for the DMI constant and A<sub>ex</sub> with R<sup>2</sup> scores of 0.991 and 0.998 respectively. The SVR model achieves R<sup>2</sup> scores of 0.927 and 0.989 for DMI constant and A<sub>ex</sub> respectively. The estimated values are in good agreement with true values, thus emphasizing the potential of ML methods for pattern recognition. |
| format | Article |
| id | doaj-art-55bc6c94ec8b496496d4abaa05018ffc |
| institution | Kabale University |
| issn | 2644-1292 |
| language | English |
| publishDate | 2024-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Open Journal of Nanotechnology |
| spelling | doaj-art-55bc6c94ec8b496496d4abaa05018ffc2025-01-24T00:02:30ZengIEEEIEEE Open Journal of Nanotechnology2644-12922024-01-01514915510.1109/OJNANO.2024.349483610748362Multitask Learning for Estimation of Magnetic Parameters Using Pattern RecognitionAnubha Sehgal0https://orcid.org/0009-0009-4218-7196Shipra Saini1https://orcid.org/0000-0002-8865-2956Hemkant Nehete2https://orcid.org/0009-0003-2367-5799Kunal Kranti Das3https://orcid.org/0009-0002-1745-6613Sourajeet Roy4https://orcid.org/0000-0002-9860-3242Brajesh Kumar Kaushik5https://orcid.org/0000-0002-6414-0032Indian Institute of Technology Roorkee, Roorkee, Uttarakhand, IndiaIndian Institute of Technology Roorkee, Roorkee, Uttarakhand, IndiaIndian Institute of Technology Roorkee, Roorkee, Uttarakhand, IndiaIndian Institute of Technology Roorkee, Roorkee, Uttarakhand, IndiaIndian Institute of Technology Roorkee, Roorkee, Uttarakhand, IndiaIndian Institute of Technology Roorkee, Roorkee, Uttarakhand, IndiaMachine learning (ML) approaches present an effective technique for accurately and efficiently predicting device parameters. Using these techniques, we introduce a multi-task convolutional neural network (CNN) model and support vector regression (SVR) model that is intended to precisely estimate two important parameters of magnetic systems such as the Dzyaloshinskii-Moriya interaction (DMI) constant and the exchange constant (A<sub>ex</sub>). The magnetic Hamiltonian encapsulates various energy components, including exchange energy, DMI, Zeeman energy, and anisotropy energy, wherein factors such as saturation magnetization, DMI strength, exchange stiffness, and anisotropy constants influence their magnitudes. Conventionally, the estimation of these parameters has been computationally intensive and time-consuming. The CNN and SVR models can simultaneously estimate both the DMI constant and the exchange constant, making it a versatile tool for magnetic system characterization. The custom CNN model performs best for the DMI constant and A<sub>ex</sub> with R<sup>2</sup> scores of 0.991 and 0.998 respectively. The SVR model achieves R<sup>2</sup> scores of 0.927 and 0.989 for DMI constant and A<sub>ex</sub> respectively. The estimated values are in good agreement with true values, thus emphasizing the potential of ML methods for pattern recognition.https://ieeexplore.ieee.org/document/10748362/Machine learningpattern recognitionmicromagneticDzyaloshinskii Moriya interactionparameter estimationexchange stiffness |
| spellingShingle | Anubha Sehgal Shipra Saini Hemkant Nehete Kunal Kranti Das Sourajeet Roy Brajesh Kumar Kaushik Multitask Learning for Estimation of Magnetic Parameters Using Pattern Recognition IEEE Open Journal of Nanotechnology Machine learning pattern recognition micromagnetic Dzyaloshinskii Moriya interaction parameter estimation exchange stiffness |
| title | Multitask Learning for Estimation of Magnetic Parameters Using Pattern Recognition |
| title_full | Multitask Learning for Estimation of Magnetic Parameters Using Pattern Recognition |
| title_fullStr | Multitask Learning for Estimation of Magnetic Parameters Using Pattern Recognition |
| title_full_unstemmed | Multitask Learning for Estimation of Magnetic Parameters Using Pattern Recognition |
| title_short | Multitask Learning for Estimation of Magnetic Parameters Using Pattern Recognition |
| title_sort | multitask learning for estimation of magnetic parameters using pattern recognition |
| topic | Machine learning pattern recognition micromagnetic Dzyaloshinskii Moriya interaction parameter estimation exchange stiffness |
| url | https://ieeexplore.ieee.org/document/10748362/ |
| work_keys_str_mv | AT anubhasehgal multitasklearningforestimationofmagneticparametersusingpatternrecognition AT shiprasaini multitasklearningforestimationofmagneticparametersusingpatternrecognition AT hemkantnehete multitasklearningforestimationofmagneticparametersusingpatternrecognition AT kunalkrantidas multitasklearningforestimationofmagneticparametersusingpatternrecognition AT sourajeetroy multitasklearningforestimationofmagneticparametersusingpatternrecognition AT brajeshkumarkaushik multitasklearningforestimationofmagneticparametersusingpatternrecognition |