| Abstrakt: |
Deep learning is successful in many fields due to its ability to learn strong feature representations without the need for hand‐crafted features, resulting in models with high representational power. However, many of these models are based on supervised learning and therefore depend on the availability of large annotated data sets. These are often difficult to obtain because they require human input. A common challenge for researchers in space weather is the sparsity of annotations in many of the available data sets, which are either unlabeled or have ambiguous labels. To alleviate the data bottleneck of loosely annotated data sets, unsupervised deep learning has become an important strategy, with anomaly detection being one of the most prominent applications. Unsupervised models have been successfully applied in various domains, such as medical imaging or video surveillance, to distinguish normal from abnormal data. In this work, we investigate how a purely unsupervised approach can be used to detect and extract solar phenomena in extreme ultraviolet images from the NASA SDO spacecraft. We show how a model based on variational autoencoders can be used to detect out‐of‐distribution samples and to localize regions of interest for solar activity. By using an unsupervised approach, we hope to contribute to space weather monitoring tools and further improve the understanding of space weather drivers. Plain Language Summary: Deep learning is a machine learning technique that extracts patterns from data without being explicitly programmed. It has been very successful in many fields because it can learn from large amounts of data and create models that excel at recognizing patterns. However, the standard deep learning method is supervised: It needs a lot of labeled data to work well. The scarcity of labeled data in the space weather domain makes it difficult to apply. To overcome this problem, we present in this article an unsupervised method that does not require labeled data. We use it to detect unusual patterns, or anomalies, in the data. In this research, we develop an unsupervised model that detects and extracts solar features from images taken by the Solar Dynamics Observatory spacecraft. The model is called a context encoding variational autoencoder. It learns to compress and reconstruct solar images. By training this model on a large set of images, we can detect unusual patterns in new images and locate regions of interest for solar activity. With this method, we hope to contribute to space weather monitoring tools and improve our understanding of the causes of space weather effects. Key Points: We introduce a way to extract a low‐dimensional feature representation from SDO images and significantly compress input imagesWe argue that unsupervised learning can support anomaly detection, thus enabling the real‐time flagging of unusual solar activityWe show how to effectively prepare and use the SDO ML version 2 data set to train a deep learning model [ABSTRACT FROM AUTHOR] |
