BZPEERReading, trust and collaboration for research
Menu

Discover

GraphFollow connections around a paper, topic or saved view.

Workspaces

HomePick up where your research flow left off.InboxHandle requests, replies and notifications from one place.CollectionsCurate reading lists, evidence sets and shareable dossiers.ResearchSee your private provenance graph, trail and imports.CollaborationMove from discovery into focused discussion rooms.PublishDraft, preview and publish full-text research articles.

Network

ExpertsFind specialists worth following or asking for help.CommunitiesJoin research circles organized around shared work.PartnersSee external organizations active around the same topics.

Opportunities

ListingsBrowse roles, calls and collaborations with clearer fit signals.

Account

Log inReturn to your reading and collaboration workspace.Create accountStart saving work, following people and joining teams.
Log inCreate account

Researcher profile

Yude Bu

Yude Bu contributes to research discovery and scholarly infrastructure.

ResearcherAffiliation not importedOpen to collaborate
astro-ph.SRastro-ph.GAastro-ph.IM

Trust snapshot

Quick read

Trust 19 - UnverifiedVerification L1Unclaimed author
5works
0followers
3topics
4close collaborators

Actions

Decide how to stay connected

Follow researcher0

Identity and collaboration

How to connect with this researcher

Claiming links this public author record to a researcher profile and unlocks direct collaboration workflows.

Log in to claim

Direct collaboration

Open a focused conversation when the fit is right

Claim this author entity first to unlock direct invitations.

Research graph

See the researcher in context

Open full explorer

Inspect adjacent work, topics, institutions and collaborators without jumping out to a separate graph page.

Building this graph slice

BZPEER is loading the nearby papers, people, topics and institutions for this page.

Published work

5 published item(s)

preprint2026arXiv

Deep learning-driven atmospheric parameter prediction for hot subdwarf stars with synthetic and observed spectra

We design a convolutional neural network (CNN) incorporating channel attention and spatial attention mechanisms to predict atmospheric parameters of hot subdwarfs. The experimental dataset comprises spectra at nine distinct signal-to-noise ratio (SNR) levels, with each SNR level containing 11 396 synthetic spectra and 945 observed spectra. The trained deep learning models achieves mean absolute errors (AME) in predicting hot subdwarf atmospheric parameters of 730 K for effective temperature (Teff ), 0.09 dex for surface gravity (log g), and 0.03 dex for helium abundance (log(nHe/nH)), respectively, which reaches the accuracy of traditional spectral fitting methods. Utilizing the trained deep learning models and low-resolution spectra from LAMOST DR12, we confirm 1512 hot subdwarfs from the catalog of hot subdwarf candidates, of which 291 are newly identified. Our results demonstrate that the deep learning model not only achieves accuracy comparable to traditional methods in obtaining hot subdwarf atmospheric parameters, but also far exceeds them in speed and efficiency, making it particularly suitable for the analysis of large datasets of hot subdwarf spectra.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

Automatic detection of low surface brightness galaxies from SDSS images

Low surface brightness (LSB) galaxies are galaxies with central surface brightness fainter than the night sky. Due to the faint nature of LSB galaxies and the comparable sky background, it is difficult to search LSB galaxies automatically and efficiently from large sky survey. In this study, we established the Low Surface Brightness Galaxies Auto Detect model (LSBG-AD), which is a data-driven model for end-to-end detection of LSB galaxies from Sloan Digital Sky Survey (SDSS) images. Object detection techniques based on deep learning are applied to the SDSS field images to identify LSB galaxies and estimate their coordinates at the same time. Applying LSBG-AD to 1120 SDSS images, we detected 1197 LSB galaxy candidates, of which 1081 samples are already known and 116 samples are newly found candidates. The B-band central surface brightness of the candidates searched by the model ranges from 22 mag arcsec $^ {- 2} $ to 24 mag arcsec $^ {- 2} $, quite consistent with the surface brightness distribution of the standard sample. 96.46\% of LSB galaxy candidates have an axis ratio ($b/a$) greater than 0.3, and 92.04\% of them have $fracDev\_r$\textless 0.4, which is also consistent with the standard sample. The results show that the LSBG-AD model learns the features of LSB galaxies of the training samples well, and can be used to search LSB galaxies without using photometric parameters. Next, this method will be used to develop efficient algorithms to detect LSB galaxies from massive images of the next generation observatories.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

Li-rich Giants Identified from LAMOST DR8 Low-Resolution Survey

A small fraction of giants possess photospheric lithium(Li) abundance higher than the value predicted by the standard stellar evolution models, and the detailed mechanisms of Li enhancement are complicated and lack a definite conclusion. In order to better understand the Li enhancement behaviors, a large and homogeneous Li-rich giants sample is needed. In this study, we designed a modified convolutional neural network model called Coord-DenseNet to determine the A(Li) of Large Sky Area Multi-Object Fiber Spectroscopic Telescope (LAMOST) low-resolution survey (LRS) giant spectra. The precision is good on the test set: MAE=0.15 dex, and σ=0.21 dex. We used this model to predict the Li abundance of more than 900,000 LAMOST DR8 LRS giant spectra and identified 7,768 Li-rich giants with Li abundances ranging from 2.0 to 5.4 dex, accounting for about 1.02% of all giants. We compared the Li abundance estimated by our work with those derived from high-resolution spectra. We found that the consistency was good if the overall deviation of 0.27 dex between them was not considered. The analysis shows that the difference is mainly due to the high A(Li) from the medium-resolution spectra in the training set. This sample of Li-rich giants dramatically expands the existing sample size of Li-rich giants and provides us with more samples to further study the formation and evolution of Li-rich giants.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

Searching for Barium Stars from the LAMOST Spectra Using the Machine Learning Method: I

Barium stars are chemically peculiar stars that exhibit enhancement of s-process elements. Chemical abundance analysis of barium stars can provide crucial clues for the study of the chemical evolution of the Galaxy. The Large Sky Area Multi-Object Fiber Spectroscopic Telescope (LAMOST) has released more than 6 million low-resolution spectra of FGK-type stars by Data Release 9 (DR9), which can significantly increase the sample size of barium stars. In this paper, we used machine learning algorithms to search for barium stars from low-resolution spectra of LAMOST. We have applied the Light Gradient Boosting Machine (LGBM) algorithm to build classifiers of barium stars based on different features, and build predictors for determining [Ba/Fe] and [Sr/Fe] of barium candidates. The classification with features in the whole spectrum performs best: for the sample with strontium enhancement, Precision = 97.81%, and Recall = 96.05%; for the sample with barium enhancement, Precision = 96.03% and Recall = 97.70%. In prediction, [Ba/Fe] estimated from BaII line at 4554 Å has smaller dispersion than that from BaII line at 4934 Å: MAE$_{4554 Å}$ = 0.07, $σ_{4554 Å}$ = 0.12. [Sr/Fe] estimated from SrII line at 4077 Å performs better than that from SrII line at 4215 Å: MAE$_{4077 Å}$ = 0.09, $σ_{4077 Å}$ = 0.16. A comparison of the LGBM and other popular algorithms shows that LGBM is accurate and efficient in classifying barium stars. This work demonstrated that machine learning can be used as an effective means to identify chemically peculiar stars and determine their elemental abundance.

0 citations0 reviewsNo code linkedOpen access
preprint2019arXiv

Searching for hot subdwarf stars from the LAMOST spectra. III. classifying the hot subdwarf stars from LAMOST DR4 using deep learning method

Hot subdwarf stars are core He burning stars located at the blue end of the horizontal branch, also known as the extreme horizontal branch. The properties of hot subdwarf stars are important for our understanding of the stellar astrophysics, globular clusters and galaxies. The spectra of hot subdwarf stars can provide us with the detailed information of the stellar atmospheric parameters (such as effective temperature, gravity, and helium abundances), which is important to clarify the astrophysical and statistical properties of hot subdwarf stars. These properties can provide important constraint on the theoretical models of hot subdwarf stars. Searching for hot subdwarf stars from the spectra data obtained by the Large Sky Area Multi-Object Fiber Spectroscopic Telescope (LAMOST) can significantly enlarge the sample size of hot subdwarf stars, and help us better study the nature of hot subdwarf stars. In this paper we study a new method of searching for hot subdwarf stars from LAMOST spectra using convolutional neural networks and support vector machine (CNN+SVM). The experiment on the spectra from LAMOST DR4 shows that CNN+SVM can classify the hot subdwarf stars accurately: the accuracy is 88.98$\%$ and the recall is 94.38 $\%$. Our research provides a new machine learning tool for searching for hot subdwarf stars in large spectroscopic surveys.

0 citations0 reviewsNo code linkedOpen access