서브메뉴
검색
Generative AI for Doppler Detection of Earth-Like Exoplanets Around Sun-Like Stars
Generative AI for Doppler Detection of Earth-Like Exoplanets Around Sun-Like Stars
상세정보
- 자료유형
- 학위논문 서양
- 최종처리일시
- 20260202103509
- ISBN
- 9798280748576
- DDC
- 530
- 저자명
- Liang, Yan.
- 서명/저자
- Generative AI for Doppler Detection of Earth-Like Exoplanets Around Sun-Like Stars
- 발행사항
- [Sl] : Princeton University, 2025
- 발행사항
- Ann Arbor : ProQuest Dissertations & Theses, 2025
- 형태사항
- 174 p
- 주기사항
- Source: Dissertations Abstracts International, Volume: 86-12, Section: B.
- 주기사항
- Advisor: Winn, Joshua;Melchior, Peter.
- 학위논문주기
- Thesis (Ph.D.)--Princeton University, 2025.
- 초록/해제
- 요약The search for Earth-like exoplanets in the habitable zones of Sun-like stars is one of the pre-eminent scientific challenges of our time. If we succeed, we will gain insights into planetary formation and the potential for life beyond the Earth. A promising path to this goal involves precise tracking of the Doppler shifts in the spectra of Sun-like stars. However, the surfaces of stars are subject to poorly understood instabilities and irregularities ("stellar activity") that interfere with precise Doppler measurements. Stellar activity is currently the main source of noise in our measurements. In this dissertation, I introduce AEstra, a deep learning method designed to distinguish the subtle Doppler shifts due to "Earth twins" from the larger distortions of spectral lines due to stellar activity. AEstra utilizes generative spectrum modeling to represent stellar activity independent of any overall Doppler shift of the spectrum, thereby isolating the planetary signal, without human supervision. I originally designed the machine-learning architecture underlying AEstra for a different purpose - to classify galaxy spectra independently of their redshifts - and used it to uncover a variety of unusual astronomical phenomena such as rare types of quasars and supernovae hosts. Modified and applied to exoplanet science, AEstra has been successful operating on simulated data and is being validated on real data obtained for the Sun. The next step will be to apply AEstra to archival collections of spectra of stars with known exoplanets. Ultimately, we will apply AEstra to newly obtained spectra from next-generation Doppler surveys such as the Terra Hunting Experiment (scheduled to begin in 2025).
- 일반주제명
- Physics
- 일반주제명
- Astrophysics
- 일반주제명
- Astronomy
- 키워드
- Exoplanet
- 키워드
- Machine learning
- 키워드
- Radial velocity
- 키워드
- Sun-like stars
- 기타저자
- Princeton University Astrophysical Sciences
- 기본자료저록
- Dissertations Abstracts International. 86-12B.
- 전자적 위치 및 접속
- 로그인 후 원문을 볼 수 있습니다.
MARC
008260126s2025 us c eng d■001000017357423
■00520260202103509
■006m o d
■007cr#unu||||||||
■020 ▼a9798280748576
■035 ▼a(MiAaPQ)AAI32003208
■040 ▼aMiAaPQ▼cMiAaPQ
■0820 ▼a530
■1001 ▼aLiang, Yan.▼0(orcid)0000-0002-1001-1235
■24510▼aGenerative AI for Doppler Detection of Earth-Like Exoplanets Around Sun-Like Stars
■260 ▼a[Sl]▼bPrinceton University▼c2025
■260 1▼aAnn Arbor▼bProQuest Dissertations & Theses▼c2025
■300 ▼a174 p
■500 ▼aSource: Dissertations Abstracts International, Volume: 86-12, Section: B.
■500 ▼aAdvisor: Winn, Joshua;Melchior, Peter.
■5021 ▼aThesis (Ph.D.)--Princeton University, 2025.
■520 ▼aThe search for Earth-like exoplanets in the habitable zones of Sun-like stars is one of the pre-eminent scientific challenges of our time. If we succeed, we will gain insights into planetary formation and the potential for life beyond the Earth. A promising path to this goal involves precise tracking of the Doppler shifts in the spectra of Sun-like stars. However, the surfaces of stars are subject to poorly understood instabilities and irregularities ("stellar activity") that interfere with precise Doppler measurements. Stellar activity is currently the main source of noise in our measurements. In this dissertation, I introduce AEstra, a deep learning method designed to distinguish the subtle Doppler shifts due to "Earth twins" from the larger distortions of spectral lines due to stellar activity. AEstra utilizes generative spectrum modeling to represent stellar activity independent of any overall Doppler shift of the spectrum, thereby isolating the planetary signal, without human supervision. I originally designed the machine-learning architecture underlying AEstra for a different purpose - to classify galaxy spectra independently of their redshifts - and used it to uncover a variety of unusual astronomical phenomena such as rare types of quasars and supernovae hosts. Modified and applied to exoplanet science, AEstra has been successful operating on simulated data and is being validated on real data obtained for the Sun. The next step will be to apply AEstra to archival collections of spectra of stars with known exoplanets. Ultimately, we will apply AEstra to newly obtained spectra from next-generation Doppler surveys such as the Terra Hunting Experiment (scheduled to begin in 2025).
■590 ▼aSchool code: 0181.
■650 4▼aPhysics
■650 4▼aAstrophysics
■650 4▼aAstronomy
■653 ▼aExoplanet
■653 ▼aMachine learning
■653 ▼aRadial velocity
■653 ▼aDoppler detection
■653 ▼aSun-like stars
■690 ▼a0605
■690 ▼a0596
■690 ▼a0800
■690 ▼a0606
■71020▼aPrinceton University▼bAstrophysical Sciences.
■7730 ▼tDissertations Abstracts International▼g86-12B.
■790 ▼a0181
■791 ▼aPh.D.
■792 ▼a2025
■793 ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17357423▼nKERIS▼z이 자료의 원문은 한국교육학술정보원에서 제공합니다.
