arXiv subjects

Sanghee Lee

Publications and source records attributed to Sanghee Lee.

6 recordsLinked to original sources

arXiv2025

Doppler imaging combined with high-cadence photometry. I. Revisiting the surface of a pre-main-sequence flare star

Latitude distribution of stellar magnetic activity is not well constrained by observations, despite its importance for a better understanding of stellar dynamos. We aim to obtain an accurate reconstruction of the surface spot distribution on the young, rapidly rotating K2 star PW And by combining spectroscopic and photometric diagnostics. In particular, we seek to assess how the inclusion of continuous high-precision TESS photometry in parallel with high-resolution spectroscopy improves latitude recovery of starspots, especially at low latitudes and in the southern hemisphere, which are poorly constrained by Doppler imaging (DI) alone. We explore the spatial origins of the observed white-light flares. We performed simultaneous Doppler imaging and light curve inversion (DI+LCI) using contemporaneous high-resolution GAOES-RV spectra from the 3.8 m Seimei telescope (R~65000) and high-precision TESS light curves. Surface reconstructions employ the SpotDIPy code to model both line profiles and continuum brightness variations. We compare DI+LCI maps with DI-only solutions, conduct artificial-spot simulations to evaluate the effects of latitude, phase coverage, and S/N on reconstruction reliability. We also investigate the spatial correlation between the DI+LCI reconstructed map and flares detected in the TESS data. The DI+LCI reconstruction reveals significant spot features at mid-to-low latitudes, equatorial regions, and even in the southern hemisphere. Simulations show that DI+LCI provides more accurate reconstructions than DI-only, especially under conditions of incomplete phase coverage and low S/N, by better recovering both spot latitudes and filling factors. A comparison between the DI+LCI map and the TESS flare timings also suggests potential association between flare occurrence and reconstructed spot longitudes.

astro-ph.SR↗

arXiv2023

Magnetic activity variability of nearby bright Sun-like stars by 4-year intensive H$α$ line monitoring

We report intensive monitoring of the activity variability in the H$α$ line for 10 Sun-like stars using the 1.88-m reflector at Okayama Branch Office, Subaru Telescope, during the last four years 2019-2022. Our aim was to investigate features of the stellar magnetic activity behaviors. We correlated the H$α$ line variability of each star with the stellar activity levels derived from the Ca II H&K line, suggesting its efficiency as a magnetic activity indicator. In analyzing the H$α$ line variation, we observed that some stars exhibited linear or quadratic trends during the observation period. Among several G- and K-type stars expected to have co-existing activity cycles, we confirmed the 2.9-yr short cycle of $ε$ Eri (K2V) from the H$α$ observations. Additionally, we established upper limits on the H$α$ variability of $β$ Com (G0V) and $κ$$^1$ Cet (G5V) concerning their expected shorter cycles. We also detected the possibility of short-term activity cycles in two F-type stars, $β$ Vir (F9V; $\sim$ 530 days) and $α$ CMi (F5IV-V; $\sim$ 130 days). The cycle in $α$ CMi was observed in only one season of our 4-yr observations, suggesting the temporal absence of the cycle period. However, for stars with planets, we did not observe significant magnetic activity variability likely associated with the planetary orbital period. It is speculated that the impact of H$α$ variability on radial velocity (RV) measurements may vary with spectral type.

astro-ph.SR↗

arXiv2023

BitE : Accelerating Learned Query Optimization in a Mixed-Workload Environment

Although the many efforts to apply deep reinforcement learning to query optimization in recent years, there remains room for improvement as query optimizers are complex entities that require hand-designed tuning of workloads and datasets. Recent research present learned query optimizations results mostly in bulks of single workloads which focus on picking up the unique traits of the specific workload. This proves to be problematic in scenarios where the different characteristics of multiple workloads and datasets are to be mixed and learned together. Henceforth, in this paper, we propose BitE, a novel ensemble learning model using database statistics and metadata to tune a learned query optimizer for enhancing performance. On the way, we introduce multiple revisions to solve several challenges: we extend the search space for the optimal Abstract SQL Plan(represented as a JSON object called ASP) by expanding hintsets, we steer the model away from the default plans that may be biased by configuring the experience with all unique plans of queries, and we deviate from the traditional loss functions and choose an alternative method to cope with underestimation and overestimation of reward. Our model achieves 19.6% more improved queries and 15.8% less regressed queries compared to the existing traditional methods whilst using a comparable level of resources.

cs.DB↗

arXiv2023

Magnetic activity variability from H$α$ line intensive monitoring for two F-type stars having a hot-Jupiter, $τ$ Bootis A and $\upsilon$ Andromedae A

We report the results of intensive monitoring of the variability in the H$α$ line for two F-type stars, $τ$ Boo and $\upsilon$ And, during the last four years 2019-2022, in order to investigate their stellar magnetic activity. The 4-year H$α$ line intensity data taken with the 1.88-m reflector at Okayama Branch Office, Subaru Telescope, shows the existence of a possible $\sim$ 123-day magnetic activity cycle of $τ$ Boo. The result of the H$α$ variability as another tracer of the magnetic activity on the chromosphere is consistent with previous studies of the Ca II H&K line and suggests that the magnetic activity cycle is persisted in $τ$ Boo. For $\upsilon$ And, we suggest a quadratic long-term trend in the H$α$ variability. Meanwhile, the short-term monitoring shows no significant period corresponding to specific variations likely induced by their hot-Jupiter in both cases ($\approx$ 3.31 and 4.62 days, respectively). In this H$α$ observation, we could not find any signature of the Star-Planet Magnetic Interaction. It is speculated that the detected magnetic activity variability of the two F-type stars is related to the stellar intrinsic dynamo.

astro-ph.SR↗

arXiv2022

Two-Stage Mesh Deep Learning for Automated Tooth Segmentation and Landmark Localization on 3D Intraoral Scans

Accurately segmenting teeth and identifying the corresponding anatomical landmarks on dental mesh models are essential in computer-aided orthodontic treatment. Manually performing these two tasks is time-consuming, tedious, and, more importantly, highly dependent on orthodontists' experiences due to the abnormality and large-scale variance of patients' teeth. Some machine learning-based methods have been designed and applied in the orthodontic field to automatically segment dental meshes (e.g., intraoral scans). In contrast, the number of studies on tooth landmark localization is still limited. This paper proposes a two-stage framework based on mesh deep learning (called TS-MDL) for joint tooth labeling and landmark identification on raw intraoral scans. Our TS-MDL first adopts an end-to-end \emph{i}MeshSegNet method (i.e., a variant of the existing MeshSegNet with both improved accuracy and efficiency) to label each tooth on the downsampled scan. Guided by the segmentation outputs, our TS-MDL further selects each tooth's region of interest (ROI) on the original mesh to construct a light-weight variant of the pioneering PointNet (i.e., PointNet-Reg) for regressing the corresponding landmark heatmaps. Our TS-MDL was evaluated on a real-clinical dataset, showing promising segmentation and localization performance. Specifically, \emph{i}MeshSegNet in the first stage of TS-MDL reached an averaged Dice similarity coefficient (DSC) at \textcolor[rgb]{0,0,0}{$0.964\pm0.054$}, significantly outperforming the original MeshSegNet. In the second stage, PointNet-Reg achieved a mean absolute error (MAE) of $0.597\pm0.761 \, mm$ in distances between the prediction and ground truth for $66$ landmarks, which is superior compared with other networks for landmark detection. All these results suggest the potential usage of our TS-MDL in orthodontics.

cs.CV↗

arXiv2020

A$^3$: Accelerating Attention Mechanisms in Neural Networks with Approximation

With the increasing computational demands of neural networks, many hardware accelerators for the neural networks have been proposed. Such existing neural network accelerators often focus on popular neural network types such as convolutional neural networks (CNNs) and recurrent neural networks (RNNs); however, not much attention has been paid to attention mechanisms, an emerging neural network primitive that enables neural networks to retrieve most relevant information from a knowledge-base, external memory, or past states. The attention mechanism is widely adopted by many state-of-the-art neural networks for computer vision, natural language processing, and machine translation, and accounts for a large portion of total execution time. We observe today's practice of implementing this mechanism using matrix-vector multiplication is suboptimal as the attention mechanism is semantically a content-based search where a large portion of computations ends up not being used. Based on this observation, we design and architect A3, which accelerates attention mechanisms in neural networks with algorithmic approximation and hardware specialization. Our proposed accelerator achieves multiple orders of magnitude improvement in energy efficiency (performance/watt) as well as substantial speedup over the state-of-the-art conventional hardware.

cs.DC↗