Data-driven models of stars

Similar documents
Making precise and accurate measurements with data-driven models

Engineering considerations for large astrophysics projects

JINA Observations, Now and in the Near Future

The Gaia-ESO Public Spectroscopic Survey a lesson for our community in use of limited telescope access. Gerry Gilmore Sofia Randich Gaia-ESO Co-PIs

Tests of MATISSE on large spectral datasets from the ESO Archive

Stars and Stellar Astrophysics. Kim Venn U. Victoria

What are the three basic types of spectra?

12. Physical Parameters from Stellar Spectra. Fundamental effective temperature calibrations Surface gravity indicators Chemical abundances

Determination of [α/fe] and its Application to SEGUE F/G Stars. Young Sun Lee

Are open clusters chemically homogeneous? Fan Liu

Outline. c.f. Zhao et al. 2006, ChJA&A, 6, 265. Stellar Abundance and Galactic Chemical Evolution through LAMOST Spectroscopic Survey

Extragalactic Astronomy

Active Galaxies and Galactic Structure Lecture 22 April 18th

Classical Methods for Determining Stellar Masses, Temperatures, and Radii

Temperature, Blackbodies & Basic Spectral Characteristics.

Milky Way s Anisotropy Profile with LAMOST/SDSS and Gaia

Galaxies & Introduction to Cosmology

Chapter 7: From theory to observations

Hierarchical Bayesian Modeling

From theory to observations

Hierarchical Bayesian Modeling of Planet Populations

Chapter 5 Light and Matter: Reading Messages from the Cosmos. How do we experience light? Colors of Light. How do light and matter interact?

Astr 323: Extragalactic Astronomy and Cosmology. Spring Quarter 2014, University of Washington, Željko Ivezić. Lecture 1:

Characterization of the exoplanet host stars. Exoplanets Properties of the host stars. Characterization of the exoplanet host stars

(Slides for Tue start here.)

LARGE QUASAR GROUPS. Kevin Rahill Astrophysics

Hubble s Law and the Cosmic Distance Scale

A Stellar Spectra 3. Stars shine at night (during the day too!). A star is a self-luminous sphere of gas. Stars are held together by gravity.

Types of Stars and the HR diagram

A100H Exploring the Universe: Discovering Galaxies. Martin D. Weinberg UMass Astronomy

DETERMINING STELLAR PARAMETERS FROM SPECTROSCOPIC OBSERVATIONS. Philip Muirhead Department of Astronomy Boston University

The Hertzsprung-Russell Diagram

(Present and) Future Surveys for Metal-Poor Stars

Astronomy 102: Stars and Galaxies Sample Review Test for Examination 3

From theory to observations

Milky Way s Mass and Stellar Halo Velocity Dispersion Profiles

Parallax: Measuring the distance to Stars

Exploring the structure and evolu4on of the Milky Way disk

Detailed chemical abundances of M-dwarf planet hosts from APOGEE observations

Stellar Astrophysics: The Classification of Stellar Spectra

Oxygen in red giants from near-infrared OH lines: 3D effects and first results from. Puerto de la Cruz, May 14, 2012! Carlos Allende Prieto!

Galac%c Unprecedented Precision (0.01 dex)

The Chemical/Dynamical Evolution of the Galactic Bulge

Dark Matter on Small Scales: Merging and Cosmogony. David W. Hogg New York University CCPP

The distance modulus in the presence of absorption is given by

A comparison of stellar atmospheric parameters from the LAMOST and APOGEE datasets

Probabilistic Catalogs and beyond...

The Apache Point Observatory Galactic Evolution Experiment. Ricardo Schiavon

Hierarchical Bayesian Modeling

ASTROPHYSICS. K D Abhyankar. Universities Press S T A R S A ND G A L A X I E S

D4.2. First release of on-line science-oriented tutorials

Milky Way Satellite Galaxies with DES

HW 5 posted. Deadline: * Monday 3.00 PM * -- Tip from the coach: Do it earlier, as practice for mid term (it covers only parts included in exam).

Introduction The Role of Astronomy p. 3 Astronomical Objects of Research p. 4 The Scale of the Universe p. 7 Spherical Astronomy Spherical

Distributed Genetic Algorithm for feature selection in Gaia RVS spectra. Application to ANN parameterization

Homework 13 (not graded; only some example ques!ons for the material from the last week or so of class)

Delicious Diameters of Dwarfs (in particular, the juicy red ones)

Hertzsprung-Russell Diagram, Flux, Luminosity, Magnitude 10 Oct

LEGA-C. The Physics of Galaxies 7 Gyr Ago. Arjen van der Wel Max Planck Institute for Astronomy, Heidelberg

Fundamental Astronomy

! p. 1. Observations. 1.1 Parameters

Structure of the Milky Way. Structure of the Milky Way. The Milky Way

arxiv: v1 [astro-ph.ga] 6 Mar 2018

Matthias Steinmetz. 16 Oct 2012 Science from the Next Generation Imaging and Spectroscopic Surveys 1

Milky Way S&G Ch 2. Milky Way in near 1 IR H-W Rixhttp://online.kitp.ucsb.edu/online/galarcheo-c15/rix/

Metal Poor Stars: A Review for Non-Observers. Charli Sakari

Future Opportunities for Collaborations: Exoplanet Astronomers & Statisticians

Unusual orbits in the Andromeda galaxy Post-16

APPLICATION FOR OBSERVING TIME

The Gaia Mission. Coryn Bailer-Jones Max Planck Institute for Astronomy Heidelberg, Germany. ISYA 2016, Tehran

Galactic, stellar (and planetary) archaeology with Gaia: The galactic white dwarf population

Properties of Stars (continued) Some Properties of Stars. What is brightness?

OPTION E, ASTROPHYSICS TEST REVIEW

The Mystery of Dark Matter

Simulating Stellar Spectra

Hertzsprung-Russell Diagram 7 Oct

Bayesian inference using Gaia data. Coryn Bailer-Jones Max Planck Institute for Astronomy, Heidelberg

The magnitude system. ASTR320 Wednesday January 30, 2019

SDSS-IV MaStar: a Large, Comprehensive, and High Quality Empirical Stellar Library

Earth-based parallax measurements have led to the conclusion that the Pleiades star cluster is about 435 light-years from Earth.

Kepler Stellar Properties Catalog Update for Q1-Q17 Transit Search


Astronomy II (ASTR-1020) Homework 2

Lecture Three: Stellar Populations. Stellar Properties: Stellar Populations = Stars in Galaxies. What defines luminous properties of galaxies

2019 Astronomy Team Selection Test

Automated analysis: SDSS, BOSS, GIRAFFE

Data-Intensive Statistical Challenges in Astrophysics

Introduction to SDSS -instruments, survey strategy, etc

A100 Exploring the Universe: The Milky Way as a Galaxy. Martin D. Weinberg UMass Astronomy

OPTION E, ASTROPHYSICS TEST REVIEW

Physical Parameters of KID , a Low-Mass, Double- Lined Eclipsing Binary

The Universe of Galaxies: from large to small. Physics of Galaxies 2012 part 1 introduction

Combining Gaia DR1, DR2 and Matthias Steinmetz (AIP) a preview on the full Gaia dataset. Matthias Steinmetz (AIP)

CHARA 2016: Adaptive Optics and Perspectives on Visible Interferometry

arxiv: v2 [astro-ph.im] 24 Jan 2018 Accepted 19 December Received 6 August 2017.

Quasars: Back to the Infant Universe

Combining probabilities

Gaia DR2 and/versus RAVE DR5: application for the semi-analytic thin disk model

Active Galactic Nuclei OIII

Stars: some basic characteristics

Transcription:

Data-driven models of stars David W. Hogg Center for Cosmology and Particle Physics, New York University Center for Data Science, New York University Max-Planck-Insitut für Astronomie, Heidelberg 2015 July 31 in collaboration with: Melissa Ness (MPIA), Hans-Walter Rix (MPIA), Anna Ho (MPIA), Gail Zasowski (JHU), and Dan Foreman-Mackey (NYU)

Annie Jump Cannon O B A F G K M temperature sequence! alphabetical order (A B F G K M O) is hydrogen-line-strength order Cannon understood the temperature sequence of stars without the benefit of physical models data-driven non-linear dimensionality reduction manifold learning (using a huge amount of prior knowledge) namesake of The Cannon

chemodynamics stars populate orbits in the Milky Way conserved actions (or chaotic equivalents) stars are formed from particular gas clouds stars have conserved surface abundances the combined action-chemical space will be far more informative than either taken independently

chemodynamics top priority for many new projects Gaia & Gaia-ESO HERMES & GALAH SDSS-III APOGEE terrifying inconsistencies in current approaches models of stars are amazingly good...... but chemical signatures are incredibly tiny

exoplanets extra-solar planets are always measured relative to their host stars you only understand a planet as well as you can understand the star

the paradox of precision astrophysics models are incredibly explanatory ΛCDM stellar spectroscopy helioseismology and yet...

the paradox of precision astrophysics models are incredibly explanatory ΛCDM stellar spectroscopy helioseismology and yet... models are wrong (ruled out) in detail χ 2 ν The χ 2 statistic is a measure of the size of your data! missing physics, approximation, computation, gastrophysics

physics-driven models put in everything you know gravity, atomic and molecular transitions, radiation make approximations to make things computable sub-grid models, mixing length, etc

machine learning the most extreme of data-driven models the data is the model none of your knowledge is relevant learn (fit) an exceedingly flexible model explain or cluster the data transformation from data to labels concept of non-parametrics concept of train, validate, and test many packages and implementations (and outrageous successes)

when does machine learning help you? train & test situation training data are statistically identical to the test data same noise amplitude same distance or redshift distribution same luminosity distribution never true! training data have accurate and precise labels therefore, we can t use vanilla machine learning! (astronomers rarely can)

data-driven models (my personal usage) make use of things you strongly believe noise model & instrument resolution causal structure (shared parameters) capitalize on huge amounts of data exceedingly flexible model concept of train, validate, and test every situation will be bespoke

label transfer for stars a few of your stars have good labels (from somewhere) can you use this to label the other stars? why would you want to do this?

label transfer for stars a few of your stars have good labels (from somewhere) can you use this to label the other stars? why would you want to do this? you don t have good models at your wavelengths? you want two surveys to be on the same system? you have some stars at high SNR, some at low SNR? you spent human time on some stars but can t on all?

stellar spectra stars are very close to black-bodies to first order, a stellar spectrum depends on effective temperature T eff and surface gravity log g

stellar spectra stars are very close to black-bodies to first order, a stellar spectrum depends on effective temperature T eff and surface gravity log g to second order, metallicity [Fe/H] and rotation

stellar spectra stars are very close to black-bodies to first order, a stellar spectrum depends on effective temperature T eff and surface gravity log g to second order, metallicity [Fe/H] and rotation to third order, tens of chemical abundances

stellar spectra all chemical information is in absorption lines corresponding to atomic and molecular transitions some 30 elements are visible in the best stars spectroscopy at is the primary tool R λ > 20, 000 λ

stellar astrophysics 1.0 1.0 0.8 0.8 0.6 1.0 A B Teff = 4750, log g = 3.0, [Fe/H] = 0.15 Teff = 4849, log g = 2.2, [Fe/H] = -1.0 normalized flux f 0.6 0.8 0.6 1.0 0.4 0.8 0.6 0.2 1.0 0.8 Teff = 3614, log g = 0.4, [Fe/H] = -0.68 Teff = 5003, log g = 2.8, [Fe/H] = -0.71 0.6 0.0 0.015200 15400 0.2 15600 158000.4 16000 16200 0.6 16400 16600 0.8 16800 1.0 wavelength λ (Å)

stellar astrophysics

SDSS-III APOGEE Galactic archaeology APOGEE DR10: 56,000 stars R = 22, 500 spectra in 1.5 < λ < 1.7 µm precise RVs and stellar parameters plan for a dozen abundances for every star (our own home-built and special continuum normalization; ask me!) APOGEE DR12: 156,000 stars now available

SDSS-III APOGEE 1.0 1.0 0.8 0.8 0.6 1.0 A B Teff = 4750, log g = 3.0, [Fe/H] = 0.15 Teff = 4849, log g = 2.2, [Fe/H] = -1.0 normalized flux f 0.6 0.8 0.6 1.0 0.4 0.8 0.6 0.2 1.0 0.8 Teff = 3614, log g = 0.4, [Fe/H] = -0.68 Teff = 5003, log g = 2.8, [Fe/H] = -0.71 0.6 0.0 0.015200 15400 0.2 15600 158000.4 16000 16200 0.6 16400 16600 0.8 16800 1.0 wavelength λ (Å)

train, validate, and test split the data into three disjoint subsets in the training step you set the parameters of your model using the training set the validation set is used to set hyperparameters or model complexity in the test step you apply the model to the test set new data to make predictions or deliver results

The Cannon: training set 543 stars (too few) from 19 clusters (too few) T eff, log g, [Fe/H] labels from APOGEE calling parameters and abundances labels slight adjustments to labels to get them onto possible isochrones terrible coverage of the main sequence only the Pleiades home-made Pleiades labels (by Ness) no [Fe/H] spread at high log g.

The Cannon: training set log g (dex) log g (dex) log g (dex) log g (dex) 0 1 2 3 4 5 0 1 2 3 4 5 0 1 2 3 4 5 0 1 2 3 4 5 5500 13 Gyr ( 2.35) 12 Gyr ( 2.33) 12.7 Gyr ( 2.06) 13 Gyr ( 1.98) 13 Gyr ( 1.78) M92 M15 M53 N5466 N4147 13 Gyr ( 1.66) 11.7 Gyr ( 1.58) 11.5 Gyr ( 1.5) 13 Gyr ( 1.33) 13 Gyr ( 003) M2 M13 M3 M5 M107 10 Gyr ( 0.82) 1 Gyr ( 0.28) 2 Gyr ( 0.20) 5 Gyr ( 0.03) 3.2 Gyr ( 0.01) M71 N2158 N2420 N188 M67 1.6 Gyr (0.02) 0.15 Gyr (+0.03) 2.5 Gyr (+0.09) 5 Gyr (+0.47) N7789 Pleiades N6819 N6791 5000 4500 4000 5500 5000 4500 4000 5500 5000 4500 4000 5500 5000 4500 4000 5500 5000 4500 4000 Teff (K) Teff (K) Teff (K) Teff (K) Teff (K)

The Cannon: training set log g (dex) log g (dex) log g (dex) log g (dex) 0 1 2 3 4 5 0 1 2 3 4 5 0 1 2 3 4 5 0 1 2 3 4 5 5500 13 Gyr ( 2.35) 12 Gyr ( 2.33) 12.7 Gyr ( 2.06) 13 Gyr ( 1.98) 13 Gyr ( 1.78) M92 M15 M53 N5466 N4147 13 Gyr ( 1.66) 11.7 Gyr ( 1.58) 11.5 Gyr ( 1.5) 13 Gyr ( 1.33) 13 Gyr ( 003) M2 M13 M3 M5 M107 10 Gyr ( 0.82) 1 Gyr ( 0.28) 2 Gyr ( 0.20) 5 Gyr ( 0.03) 3.2 Gyr ( 0.01) M71 N2158 N2420 N188 M67 1.6 Gyr (0.02) 0.15 Gyr (+0.03) 2.5 Gyr (+0.09) 5 Gyr (+0.47) N7789 Pleiades N6819 N6791 5000 4500 4000 5500 5000 4500 4000 5500 5000 4500 4000 5500 5000 4500 4000 5500 5000 4500 4000 Teff (K) Teff (K) Teff (K) Teff (K) Teff (K)

The Cannon: model a generative model of the APOGEE spectra given label vector l, predict flux vector f probabilistic prediction p(f l, θ) use every spectral pixel s uncertainty variance σλn 2 responsibly details: spectral expectation is quadratic in the labels every wavelength λ treated independently an intrinsic Gaussian scatter s 2 λ at every wavelength λ 80,000 free parameters in θ!

The Cannon: model ln p(f n l n, θ) = L ln p(f λn l n, θ λ, s 2 λ) λ=1 ln p(f λn l n, θ λ, s 2 λ) = 1 [f λn θ λt l n] 2 2 σλn 2 + + ln(σ s2 λn 2 + s 2 λ) λ l T {1, T eff, log g, [Fe/H], Teff, 2 T eff log g,, [Fe/H] 2} θ T { θ λ, s 2 } L λ λ=1

The Cannon: model ln p(f n l n, θ) training step: optimize w.r.t. parameters θ at fixed labels l using training-set data linear least squares every wavelength λ treated independently test step: optimize w.r.t. labels l at fixed parameters θ using test-set (survey) data non-linear optimization every star treated independently

The Cannon: model training

The Cannon: model training

The Cannon: model training cross-validation

The Cannon: results The Cannon is far faster than physical modeling model trains in seconds (thousands of fits) The Cannon labels all 56,000 stars in APOGEE DR10 in two hours (pure Python on a laptop) labels appear sensible The Cannon labels lie near sensible isochrones scatter against APOGEE labels consistent with APOGEE precision successfully puts labels on dwarfs

The Cannon: test time

The Cannon: test time

The Cannon: test time

The Cannon: test time

The Cannon: test time

The Cannon: test time

The Cannon: comparison with APOGEE labels

The Cannon: label veracity

The Cannon: label veracity

The Cannon: works at low signal-to-noise

The Cannon: works at low signal-to-noise

The Cannon: results The Cannon is far faster than physical modeling model trains in seconds (thousands of fits) The Cannon labels all 56,000 stars in APOGEE DR10 in two hours (pure Python on a laptop) labels appear sensible The Cannon labels lie near sensible isochrones scatter against APOGEE labels consistent with APOGEE precision successfully puts labels on dwarfs

The Cannon: label transfer from APOGEE to LAMOST

The Cannon: shortcuts and choices no Bayes; no partial or noisy labels quadratic order replacing polynomial with a Gaussian process continuous model complexity; non-parametric spectral representation too-small training set only three labels age, [α/fe] splitting the giant branch how to go to many elements?

The Cannon: masses and ages for red giants

The Cannon: masses and ages for red giants

applications for data-driven models Kepler and K2 light curves the first systematic exoplanet catalog from K2 data Foreman-Mackey et al. (arxiv:1502.04715) building a consistent all-sky stellar parameter system for Gaia quasar target selection XDQSO and XDQSOz Bovy et al., 2011 (ApJ 729 141), 2012 (ApJ 749 41) CMB foregrounds

data-driven models incredibly powerful tools clustering, label transfer, prediction, de-noising make use of things you strongly believe especially the noise model every situation will be bespoke expect to get dirty