EDEkaterina Dineva

Ekaterina Dineva

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Ekaterina Dineva has a total of 77 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Probabilistic Solar Flare Forecasting via Weakly Supervised Contrastive Refinement of VAE Latent Spaces

    by Ekatarina Dineva, Jasmina Magdalenic, George Miloshevich, Panagiotis Gonidakis, Francesco Carella, Stefaan Poedts

    Reliable solar flare forecasting is limited by two forms of class imbalance in active region time series: (i) the overwhelming dominance of the non-flaring, quiet state over the eruptive state, and (ii) the insufficient separability between common, physically similar event classes (e.g. C-class versus M-class flares). Although empirical parameters derived from the photospheric vector magnetic field (VMF), such as those provided by SDO/HMI SHARP products, capture aspects of active region complexity and free energy buildup, they often evolve smoothly and overlap across flare classes. Consequently, while many models can distinguish between flares and no-flares reasonably well, they struggle to distinguish flare magnitude and association with eruptive phenomena (e.g. CMEs) using photospheric information alone. This suggests that improved flare-class separation requires (a) the explicit definition of what constitutes 'similarity' between pre-flare states, and (b) parametrization that emphasizes flare-relevant structure over common active region features.We investigate a representation learning strategy that combines the parametrization of SDO/HMI SHARP VMF cutouts using a Variational Autoencoder (VAE) with a contrastive stage to reshape the resulting embedding geometry. First, a VAE is trained to encode SHARP cutouts into compact latent vectors that capture active region morphology. These vectors are then refined using a Siamese-like objective constructed from weak supervision, which uses event labels and empirical SHARP parameters as proxies for elevated flare likelihood. The contrastive stage then uses this weak supervision to encourage a latent geometry that better reflects flare-relevant evolution. This study emphasizes latent-space structure, i.e. neighborhood consistency and class-conditional clustering, and evaluates whether these properties facilitate improved probabilistic prediction across multiple forecast horizons, by training lightweight downstream models on (i) empirical parameters, (ii) VAE latents and (iii) their combined representations.

  • Open Access

    Prediction of Time-evolving Radial Magnetic Fields on the Solar Surface Using Deep Learning

    by Hyun-Jin Jeong, Stefaan Poedts, Haopeng Wang, Ekatarina Dineva, Panagiotis Gonidakis, Francesco Carella, George Miloshevich, Luis Linan, Harim Lee


  • Open Access

    Parametrization of SHARP Vector Magnetic Field Using Disentangled Representation Learning

    by Ekatarina Dineva, George Miloshevich, Giovanni Lapenta, Jasmina Magdalenic Zhukov, Stefaan Poedts

    The rapid growth of high-dimensional data in solar physics presents significant challenges for analysis and interpretation, making it an excellent domain for the application of machine learning (ML) algorithms. Synoptic full-disk observations with the Solar Dynamics Observatory (SDO) provide continuous observations of the solar magnetic activity over more than one solar cycle, facilitating the study of solar variability and space weather impacts. The Space-weather HMI Active Region Patches (SHARP) vector magnetic field (VMF) maps and parameters, based on Helioseismic and Magnetic Imager (HMI) full-disk observations, are developed to study the magnetic evolution of individual active regions and flare triggering mechanisms. We present a method for active region parametrization by combining empirical parameters and ML-extracted features. Time series of SHARP VMF maps are used as input for the Disentangled Variational Autoencoder (VAE), a Disentangled Representation Learning (DRL) algorithm that facilitates the extraction of a low-dimensional feature representation. The VAE model is used to encode generalized information about nonlinear dynamical systems, i.e., a solar active region, aiming to isolate distinct factors of variation in the data, allowing a clearer interpretation of physical processes. We demonstrate how the ML features can be used to identify and study the stages of the magnetic patches evolution. These are benchmarked with SHARP parameters, relating empirical and learned features. Furthermore, the empirical dataset enhanced with ML features can be used to analyze the development of individual active regions and searching for eruption precursors.

  • Open Access

    Machine Learning Algorithms for Autonomous Space Mission Operations

    by Tommaso Torda, Tommaso Alberti, Giuseppe Consolini, Rossana De Marco, Ekaterina Dineva, Jonah Ekelund, Panagiotis Gonidakis, Monica Laurenza, Maria Federica Marcucci, Stefano Markidis, et al.

    The Automatics in SpAce exploration (ASAP) project has as a goal the design and development of Machine Learning algorithms for the automation of operations to be implemented on the on-board processors of space missions. In the framework of ASAP a set of ML algorithms for on-board science operations of space missions have been developed/optimized on consumer-grade computing systems to be further selected for orting of existent ML models directly on an FPGA prototype. In more detail, algorithms pertaining to four main use cases have been considered: the autonomous triggering of special measurement modes and the selective downlink of plasma environment parameters; the advanced on-board data analysis of three-dimensional particle distribution functions; the on-board analysis of solar images; the on-board prediction capability of SEP related hazards. Here we describe the algorithms, their performances and requirements for the on-board implementation. ASAP has received funding from the EU’s HORIZON Research and Innovation Action (GA no.101082633)

  • Open Access

    Bypassing the static input size of neural networks in flare forecasting by using spatial pyramid pooling

    by P. Vong, L. R. Dolla, A. Koukras, J. Gustin, J. Amaya, E. Dineva, G. Lapenta

    Context. The spatial extension of active regions of the Sun (hence their associated images) can strongly vary from one case to the next. This inhomogeneity is a problem when using convolutional neural networks (CNNs) to study solar flares, as they generally use input images of a fixed size. Different processes can be performed to retrieve a database with homogeneous-sized images, such as coarse resizing, cropping, or padding of raw images. Unfortunately, key features can be lost or distorted beyond recognition during these processes. This can lead to a deterioration of the ability of CNNs to classify flares of different soft X-ray classes, especially those from active regions with structures of great complexity. Aims. This study aims to implement and test a CNN architecture that retains features of characteristic scales as fine as the original resolution of the input images. Methods. We compared the performance of two CNN architectures for solar flare prediction. The first one is a traditional CNN with convolution layers, batch normalization layers, max pooling layers, and resized input, whereas the other implements a spatial pyramid pooling (SPP) layer instead of a max pooling layer before the flattening layer and without any input resizing. Both were trained on the Spaceweather HMI Active Region Patch (SHARP) line of sight magnetogram database, which was generated from data collected by the Helioseismic and Magnetic Imager on board the Solar Dynamics Observatory from May 2010 to August 2021 only using images within 45°of the central meridian of the Sun. We also studied two cases of binary classification. In the first case, our model had to distinguish active regions producing flares in less than 24 h of class ≥C1.0 from active regions producing flares in more than 24 h or never. In the second case, it had to distinguish active regions producing flares in less than 24 h of class ≥M1.0 from active regions producing flares in more than 24 h or never, or flares in less than 24 h but of class < M1.0. The impact of the use of a score-oriented loss (SOL) function optimizing the true skill statistics (TSS) metric instead of a binary cross-entropy (BCE) loss function is also studied and discussed in this work. Results. Our models implementing an SPP layer and trained using a BCE loss function outperform the traditional CNN models, with an average increase of 0.1 in TSS and 0.17 in precision metrics when predicting flares ≥C1.0 within 24 h. However, their performances degrade sharply along the other models studied in this paper when trained to classify images of ≥M1.0 flares. Conclusions. We prove the higher efficiency of a CNN model that includes an SPP layer in predicting solar flares. The degradation of prediction performance of this model when the images of active regions producing a C class flare are classified as negative may be attributed to its success in identifying features that appear in active regions only a few hours before the flare, independent of their soft X-ray class. The development of explainable artificial intelligence tools adapted to this architecture in future projects will be interesting for the study of solar flare-triggering mechanisms.


  • Open Access

    Investigating the absence of stellar CMEs through solar observations 

    by Alexander G.M. Pietrow, Michael Cretignier, Malcolm K. Druett, Julian D. Alvarado-Gómez, Stefan J. Hofmeister, Meetu Verma, Robert Kamlah, Martina Barlatella, Eliana M. Amazo-Gómez, Ioannis Kontogiannis, et al.

    Coronal Mass Ejections (CMEs) remain a focal point of solar and stellar research due to their significant impact on space weather dynamics and exoplanet habitability. Unfortunately, it has so far proven difficult to measure these events on other stars, with only a handful of confirmed detections. On the Sun, strong flares (X1-class and above) are almost always accompanied by a CME. This connection has not been found for other stars, where strong flares are regularly detected without a CME counterpart. To investigate this discrepancy we compare resolved solar observations taken from the Swedish 1-m Solar Telescope with disk-integrated Sun-as-a-star observations taken from the HARPS-N solar telescope. We studied two strong X-class flares, one of which was accompanied by a large (halo)CME, and one that was not. While we have successfully detected flare-related signatures in the activity indices and the radial velocity profile in the Sun-as-a-star data, we detect no significant differences between the two flares and no indications of the presence of the CME, despite other works having previously detected CMEs in Sun-as-a-star data.We propose that the absence of CME signatures in our data is due to a geometric effect. CMEs far enough away from the disk center are likely to be oriented in such a way that they have limited line-of-sight velocity, and thus cannot produce a strong enough Doppler signature. Therefore, we believe that the lack of observed stellar CMEs is at least partly an observational limitation and does not necessarily represent the underlying physical reality.

  • Open Access

    Solar active region evolution and imminent flaring activity through color-coded visualization of photospheric vector magnetograms

    by I. Kontogiannis, A. G. M. Pietrow, M. K. Druett, E. Dineva, M. Verma, C. Denker

    Context. The emergence of magnetic flux, its transition to complex configurations, and the pre-eruptive state of active regions are probed using photospheric magnetograms. Aims. Our aim is to pinpoint different evolutionary stages in emerging active regions, explore their differences, and produce parameters that could advance flare prediction using color-coded maps of the photospheric magnetic field. Methods. The three components of the photospheric magnetic field vector are combined to create color-combined magnetograms (COCOMAGs). From these, the areas occupied by different color hues are extracted, creating appropriate time series (color curves). These COCOMAGs and color curves are used as proxies of the active region evolution and its complexity. Results. The morphology of COCOMAGs showcases typical features of active regions, such as sunspots, plages, and sheared polarity inversion lines. The color curves represent the area occupied by photospheric magnetic field of different orientation and contain information pertaining to the evolutionary stages of active regions. During emergence, most of the region area is dominated by horizontal or highly inclined magnetic field, which is gradually replaced by more vertical magnetic field. In complex regions, large parts are covered by highly inclined magnetic fields, appearing as abrupt color changes in COCOMAGs. The decay of a region is signified by a domination of vertical magnetic field, indicating a gradual relaxation of the magnetic field configuration. The color curves exhibit a varying degree of correlation with active region complexity. Particularly the red and magenta color curves, which represent strong, purely horizontal magnetic field, are good indicators of future flaring activity. Conclusions. Color-combined magnetograms facilitate a comprehensive view of the evolution of active regions and their complexity. They offer a framework for the treatment of complex observations and can be used in pattern recognition, feature extraction, and flare-prediction schemes.


  • High-Resolution Sun-as-a-Star Spectroscopy of the Partial Solar Eclipse of 2017 August 21

    by Ekaterina Dineva, Carsten Denker, Meetu Verma, Klaus Strassmeier, Ilya Ilyin, Ivan Milic

  • Open Access

    Comparative study of nearly‐grazing and fully‐grazing exoplanet system parameters derived with TESS and ground‐based instruments

    by Xanthippi Alexoudi, Ekaterina Dineva, Sydney Barnes, Klaus G. Strassmeier

    Grazing transiting systems are rare and difficult to parameterize due to their partial transit geometry. We investigated a sample of 43 nearly‐grazing to fully‐grazing transiting exoplanetary systems, using datasets provided by the Transiting Exoplanet Satellite Survey (TESS), aiming first to refine major system parameters. We focus on the two parameters, impact parameter and planetary radius , through a comparison of TESS light curves with ground‐based light curves. We investigate if there is a potential systematic trend between ground‐based and space‐based investigations that would possibly lead to a redefinition of the grazing nature of those exoplanetary systems. Our results have shown that the value is confirmed for most of the systems within , and refined significantly for Kilodegree Extremely Little Telescope (KELT)‐8 b, while the uncertainty on was improved for five systems with . The measurements of four systems were successfully refined, while the uncertainty of was improved for the majority of exoplanets with . Furthermore, a systematic trend is more pronounced for systems with , where there is a mean overestimation on the ground‐based derived by . We concluded eventually that from our sample only five systems are truly grazing, while four systems that were considered as grazing are not.

  • Open Access

    Improved High-resolution Fast Imager

    by Carsten J. Denker, Meetu Verma, Aneta Wiśniewska, Robert Kamlah, Ioannis Kontogiannis, Ekaterina Dineva, Jürgen Rendtel, Svend-Marian Bauer, Mario Dionies, Hakan Önel, et al.

    The improved High-resolution Fast Imager (HiFI+) is a multiwavelength imaging filtergraph, which was commissioned at the GREGOR solar telescope at Observatorio del Teide, Izaña, Tenerife, Spain, in March 2022 – followed by science verification in April 2022, after which it entered routine observations. Three camera control computers with two synchronized sCMOS and CMOS cameras each provide near diffraction-limited imaging at high cadence in six wavelength bands (Ca ii H at 396.8 nm, G-band at 430.7 nm, blue continuum at 450.6 nm, narrow- and broad-band Hα at 656.3 nm, and TiO bandhead at 705.8 nm). This unique combination of photospheric and chromospheric images provides “tomographic” access to the dynamic Sun and complements spectropolarimetric observations at the GREGOR telescope. High image acquisition rates of 50 and 100 Hz facilitate image restoration, where time series of restored images have a typical cadence of 6 and 12 s, which is sufficient to resolve the dynamics of the solar photosphere and chromosphere. In principle, all imaging channels can be restored individually using the speckle masking technique or multiframe blind deconvolution (MFBD). However, images recorded strictly simultaneously in the narrow-/broad-band Hα and the G-band/blue continuum channels can be pairwise subjected to multiobject multiframe deconvolution (MOMFBD) expanding the science capabilities of HiFI+. For example, the narrow-band (FWHM = 60 nm) Halle Hα Lyot filter isolates the Hα line core, which facilitates matching chromospheric fibrils and filamentary structures to photospheric bright points. Likewise, dividing G-band by blue continuum images enhances small-scale brightenings, which are often related to small-scale magnetic fields so that their evolution can be tracked in time. A detailed description of the improved high-cadence, large-format imaging system is presented and its performance is assessed based on first-light observations.