Journal Article10.1007/s41116-023-00038-x
Machine learning in solar physics
TL;DR: Using techniques such as deep learning, the use of machine learning can help to automate the analysis of solar data, reducing the need for manual labor and increasing the efficiency of research in this field.
read more
About: This article is published in Living Reviews in Solar Physics. The article was published on 27 Jun 2023. The article focuses on the topics: Computer science.
read more
Chat with Paper
AI Agents for this Paper
Find similar papers on Google Scholar, PubMed and Arxiv
Write a critical review of this paper
Analyze citations of this paper to find unaddressed research gaps
Figures
Fig. 2 Schematic representation of supervised learning, with the feature space X and the target space Y . The aim is to define or learn mappings f and g between the two spaces by taking advantage of the information encoded in the pairwise relations were known in advance for a specific sample of the two spaces. 
Fig. 10 Space Weather HMI Active Region Patches (SHARPs) identified by a computer tracking algorithm. In this image, two SHARPs have been identified and are marked by rectangular bounding boxes. 
Fig. 9 SPectral classifications of an IBIS observation, where the color bar relates to the spectral shape classified, with ‘0’ and ‘4’ representing pure absorption and emission profiles, respectively. The umbra/penumbra boundary is highlighted using a black contour (from MacBride et al., 2021) 
Table 1 Contingency table for binary classification. We denote TP, TN, FP and FN as the number of true positives, true negatives, false positives and false negatives, respectively. 
Table 2 Evaluation metrics for binary classification. Refer to Table 1 for the definition of classes. In the definition for the Gilbert Skill Score, CH (chance hits) is the Accuracy for a random forecast model. The probability that a random forecast outputs a positive is uncorrelated with the underlying probability of the event. Hence, the joint probability for the Accuracy can be factored, giving CH = P(Forecast : Yes)×P(Event : Yes) = (TP+FP) 
Fig. 4 Building block of a fully-connected neural network. Each input of the previous layer is connected to each neuron of the output. Each connection is represented by different lines where the width is proportional to the absolute value of the weight. Solid lines represent positive weights while dashed lines refer to negative weights.
Citations
An Improved Prediction of Solar Cycles 25 and 26 Using the Informer Model: Gnevyshev Peaks and North–South Asymmetry
Jie Cao,Tingting Xu,Linhua Deng,Xueliang Zhou,Shangxi Li,Y B Liu,Wen-hua Wang,Weihong Zhou +7 more
TL;DR: This study uses the Informer model to predict Solar Cycles 25 and 26, forecasting weak-moderate cycles with stronger amplitudes than Cycle 24, and significant north-south asymmetry, with the southern hemisphere exhibiting stronger activity during Cycle 25.
4
Predicting Solar Cycle 26 Using the Polar Flux as a Precursor, Spectral Analysis, and Machine Learning: Crossing a Gleissberg Minimum?
José‐Víctor Rodríguez,V. M. S. Carrasco,Ignacio Rodríguez‐Rodríguez,A. J. P. Aparicio,J. M. Vaquero +4 more
TL;DR: This study predicts Solar Cycle 26's sunspot number using machine learning, polar flux, and spectral analysis, forecasting a maximum peak of 121.2 in September 2034, with Cycle 25's peak below average, suggesting a Gleissberg cycle minimum.
2
Hybrid Model of Natural Time Series with Neural Network Component and Adaptive Nonlinear Scheme: Application for Anomaly Detection
Oksana Mandrikova,Bogdana Mandrikova +1 more
TL;DR: This investigation showed that the developed HMTS adequately describes neutron monitor data and has satisfactory results from the point of view of numeric performance, and results show that the developed HMTS has the potential to address the problem of anomaly detection in neutron monitor data even when the number of operating stations is small.
1
Short-term solar eruptive activity prediction models based on machine learning approaches: A review
Xin Huang,Zhongrui Zhao,Yufeng Zhong,Long Xu,Marianna B. Korsós,R. Erdélyi +5 more
1
References
Deep Residual Learning for Image Recognition
Kaiming He,Xiangyu Zhang,Shaoqing Ren,Jian Sun +3 more
- 27 Jun 2016
TL;DR: In this article, the authors proposed a residual learning framework to ease the training of networks that are substantially deeper than those used previously, which won the 1st place on the ILSVRC 2015 classification task.
•Proceedings Article
Adam: A Method for Stochastic Optimization
Diederik P. Kingma,Jimmy Ba +1 more
- 01 Jan 2015
TL;DR: This work introduces Adam, an algorithm for first-order gradient-based optimization of stochastic objective functions, based on adaptive estimates of lower-order moments, and provides a regret bound on the convergence rate that is comparable to the best known results under the online convex optimization framework.
138.5K
•Posted Content
Deep Residual Learning for Image Recognition
TL;DR: This work presents a residual learning framework to ease the training of networks that are substantially deeper than those used previously, and provides comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth.
117.9K
•Proceedings Article
Very Deep Convolutional Networks for Large-Scale Image Recognition
Karen Simonyan,Andrew Zisserman +1 more
- 04 Sep 2014
TL;DR: This work investigates the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting using an architecture with very small convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers.
102.6K
Long short-term memory
TL;DR: A novel, efficient, gradient based method called long short-term memory (LSTM) is introduced, which can learn to bridge minimal time lags in excess of 1000 discrete-time steps by enforcing constant error flow through constant error carousels within special units.
99K