Big Data Inference. Combining Hierarchical Bayes and Machine Learning to Improve Photometric Redshifts. Josh Speagle 1,
|
|
- Judith McBride
- 5 years ago
- Views:
Transcription
1 ICHASC, Apr Big Data Inference Combining Hierarchical Bayes and Machine Learning to Improve Photometric Redshifts Josh Speagle 1, In collaboration with: Boris Leistedt 2, Alexie Leauthaud 3, Daniel Eisenstein 1, Kevin Bundy 3,4, Peter Capak 5, Ben Hoyle 6, Daniel Masters 5, Daniel Mortlock 7, and Hiranya Peiris 8,9 NSF Graduate Research Fellow, 1 Harvard University, 2 NYU, 3 UC Santa Cruz, 4 UC Observatories, 5 IPAC, 6 LMU Munich, 7 Imperial College London, 8 University College London, 9 Oskar Klein Centre for Cosmoparticle Physics
2 What are Photometric Redshifts?
3 Basic Concept?
4 Basic Concept? SDSS Skyserver Spectrum
5 Basic Concept Spec-z SDSS Skyserver Spectrum
6 Basic Concept Spec-z SDSS Skyserver Spectrum
7 Basic Concept Spec-z SDSS Skyserver Spectrum
8 Basic Concept? SDSS Skyserver Broadband filters
9 Basic Concept Spectral Energy Distribution (SED)? SDSS Skyserver Broadband filters
10 Basic Concept Spectral Energy Distribution (SED) SDSS Skyserver Broadband filters
11
12 Why do we care?
13 Because we have to. Why do we care?
14 Because we have to. Why do we care? Many questions now require large samples of galaxies to answer now entering the Big Data era of astronomy.
15 Because we have to. Why do we care? Many questions now require large samples of galaxies to answer now entering the Big Data era of astronomy. Wide-field imaging surveys much cheaper and faster than spectroscopic surveys. Also can see fainter objects.
16 Because we have to. Why do we care? Many questions now require large samples of galaxies to answer now entering the Big Data era of astronomy. Wide-field imaging surveys much cheaper and faster than spectroscopic surveys. Also can see fainter objects. ~100x increase in sample size, diversity makes up for photo-z uncertainties. (Detailed studies can rely on ~1% spectroscopic subsample.)
17 Science Case Precision cosmology Using large samples of galaxies to pin down the dark energy equation of state, growth of large-scale structure, etc. Taken from
18 Computing Photo-z s
19 Photo-z s: Statistically Speaking Star formation rate Star formation history Redshift Stellar mass AGN activity Dust content Metallicity
20 Photo-z s: Statistically Speaking A forward modeling problem: can we construct a model from parameters we care about that matches the observed SED? Star formation rate Star formation history Redshift Stellar mass AGN activity Dust content Metallicity
21 Flux (Jy; normalized) Photometry = F( a T gal ) Traditional Approach e.g., COSMOS: Ilbert et al. (2009), Laigle et al. (2016) Galaxy Wavelength (A)
22 Flux (Jy; normalized) Traditional Approach e.g., COSMOS: Ilbert et al. (2009), Laigle et al. (2016) Photometry = F( a, b T gal, T em ) Emission lines Galaxy Wavelength (A)
23 Flux (Jy; normalized) Traditional Approach e.g., COSMOS: Ilbert et al. (2009), Laigle et al. (2016) Photometry = F( a, b, c, E(B V) T gal, T em, T dust ) Emission lines Galaxy Reddening Wavelength (A)
24 Flux (Jy; normalized) Traditional Approach e.g., COSMOS: Ilbert et al. (2009), Laigle et al. (2016) Photometry = F( a, b, c, E(B V), z T gal, T em, T dust ) Emission lines Galaxy Reddening Redshift Wavelength (A)
25 Flux (Jy; normalized) Traditional Approach e.g., COSMOS: Ilbert et al. (2009), Laigle et al. (2016) Photometry = F( a, b, c, E(B V), z T gal, T em, T dust, T IGM ) Emission lines Galaxy Reddening Redshift Lyman limit IGM Lyα Wavelength (A)
26 Flux (Jy; normalized) Traditional Approach e.g., COSMOS: Ilbert et al. (2009), Laigle et al. (2016) Photometry = F( a, b, c, E(B V), z T gal, T em, T dust, T IGM, filters) Emission lines Galaxy Reddening Redshift Photometry IGM Wavelength (A)
27 Flux (Jy; normalized) Traditional Approach e.g., COSMOS: Ilbert et al. (2009), Laigle et al. (2016) Photometry = F( a, b, c, E(B V), z T gal, T em, T dust, T IGM, filters) Emission lines Galaxy Reddening Redshift Photometry IGM Wavelength (A)
28 Photo-z s: Statistically Speaking A forward modeling problem: can we construct a model from parameters we care about that matches the observed SED? Star formation rate Star formation history Redshift Stellar mass AGN activity Dust content Metallicity
29 Photo-z s: Statistically Speaking An inverse mapping problem: can we use machine learning to construct a mapping from color to redshift? Star formation rate Star formation history Redshift Stellar mass AGN activity Dust content Metallicity
30 Machine Learning Training set
31 Machine Learning Decision Tree
32 Machine Learning
33 Machine Learning z 1 z 2 z 3 z 4 z 5 z 6 z 7 z 8 Target (redshift)
34 Machine Learning ~ P C g Feature vector (observed color) z 1 z 2 z 3 z 4 z 5 z 6 z 7 z 8 Target (redshift)
35 Machine Learning ~ P C g Feature vector (observed color) z 1 z 2 z 3 z 4 z 5 z 6 z 7 z 8 Target (redshift)
36 Machine Learning
37 Machine Learning z 1 z 2 z 3 z 4 z 5 z 6 z 7 z Target Weight Kernel Density Estimation P z g
38 Two Communities Model-fitting approaches (color-redshift relation assumed) Machine-learning approaches (feature-redshift relation derived)
39 Two Communities Model-fitting approaches (color-redshift relation assumed) Machine-learning approaches (feature-redshift relation derived)
40 Two Communities Model-fitting approaches (color-redshift relation assumed) Probabilistic Interpretable Sensitive to systematics Generally slow Machine-learning approaches (feature-redshift relation derived)
41 Two Communities Model-fitting approaches (color-redshift relation assumed) Probabilistic Interpretable Sensitive to systematics Generally slow Flexible, data-driven More robust to systematics Generally fast Difficult to interpret Difficult to derive PDFs Machine-learning approaches (feature-redshift relation derived)
42 What are we doing?
43 Splitting Photo-z s into Two Parts Mapping from features to redshift. Can be done through model fitting and/or machine learning. P z F or P F z, η
44 Splitting Photo-z s into Two Parts Mapping from features to redshift. Can be done through model fitting and/or machine learning. P z F or P F z, η Propagating uncertainties in features. P F F, C
45 Splitting Photo-z s into Two Parts Mapping from features to redshift. Can be done through model fitting and/or machine learning. P z F h, C h P F h F h, C h Propagating uncertainties in features. P F g F g, C g
46 Splitting Photo-z s into Two Parts Mapping from features to redshift. Can be done through model fitting and/or machine learning. P z F h, C h P F h F h, C h Propagating uncertainties in features. P F g F g, C g
47 Splitting Photo-z s into Two Parts Mapping from features to redshift. Can be done through model fitting and/or machine learning. P z F h, C h P F h F h, C h Propagating uncertainties in features. P F g F g, C g
48 Feature Projection P F h F h, C h P F g F g, C g
49 The Big Data Assumption P z g = h P z h P h g
50 The Big Data Assumption redshift P z g = target galaxy g h redshift PDF P z h P h g training galaxy h photometric posterior
51 The Big Data Assumption P z g = P z h P h g h = h P z h P g h P h P g
52 The Big Data Assumption P z g = P z h P h g h = P z h 1. What is this likelihood? h P g h P h P g
53 The Big Data Assumption P z g = P z h P h g = h h P z h 1. What is this likelihood? 2. How do we compute it? P g h P h P g
54 The Big Data Assumption P z g = = h h P z h P z h P h g band mask P g h, b P h P g 1. What is this likelihood? 2. How do we compute it? 3. How do we deal with missing data?
55 The Big Data Assumption P z g = h h P z h P h g, s g = 1, S g P z h P g h, b P h P g galaxy is observed selection effects 1. What is this likelihood? 2. How do we compute it? 3. How do we deal with missing data? 4. How do we incorporate selection effects?
56 The Big Data Assumption P z g = P z h P h g, s g = 1, S g h h P z h P g h, b P h P g 1. What is this likelihood? 2. How do we compute it? 3. How do we deal with missing data? 4. How do we incorporate selection effects? 5. What is our prior?
57 The Big Data Assumption P z g = P z h P h g, s g = 1, S g h h P z h P g h, b P h P g 1. What is this likelihood? 2. How do we compute it? 3. How do we deal with missing data? 4. How do we incorporate selection effects? 5. What is our prior?
58 Likelihood / Distance Metric What metric(s) to use? P g h Training galaxy Observed galaxy
59 Likelihood / Distance Metric What metric(s) to use? P g h Training galaxy Observed galaxy NGC 4414 NGC 5457
60 Likelihood / Distance Metric What metric(s) to use? P g h Training galaxy SDSS Skyserver Observed galaxy Broadband filters
61 Likelihood / Distance Metric What metric(s) to use? P g h Training galaxy SDSS Skyserver Observed galaxy Assume data is normally distributed. Broadband filters P F g F g, C g N F g F g, C g
62 Likelihood / Distance Metric What metric(s) to use? P g h Color space (traditional) χ 2 = i F g,i s F h,i 2 σ 2 g,i + s 2 2 σ h,i
63 Likelihood / Distance Metric What metric(s) to use? P g h Color space (traditional) χ 2 = i F g,i s F h,i 2 σ 2 g,i + s 2 2 σ h,i
64 Likelihood / Distance Metric What metric(s) to use? P g h Color space (traditional) χ 2 = i F g,i s F h,i 2 σ 2 g,i + s 2 2 σ h,i Magnitude space ( new ) χ 2 = i F g,i F h,i 2 σ 2 2 g,i + σ h,i
65 Likelihood / Distance Metric What metric(s) to use? P g h Color space (traditional) Requires magnitude priors to account for galaxy evolution. χ 2 = i F g,i s F h,i 2 σ 2 g,i + s 2 2 σ h,i Scale-free Magnitude space ( new ) Requires good sampling in full magnitude space. χ 2 = i F g,i F h,i 2 σ 2 2 g,i + σ h,i Scale-dependent
66 Population: Mag v Color
67 The Big Data Assumption P z g = P z h P h g, s g = 1, S g h h P z h P g h, b P h P g 1. What is this likelihood? 2. How do we compute it? 3. How do we deal with missing data? 4. How do we incorporate selection effects? 5. What is our prior?
68 The Big Data Assumption P z g = P z h P h g, s g = 1, S g h h P z h P g h, b P h P g 1. What is this likelihood? 2. How do we compute it? 3. How do we deal with missing data? 4. How do we incorporate selection effects? 5. What is our prior?
69 The Problem Training data Galaxy Likelihood P h g Photometric training set Cool et al. (2007)
70 Machine Learning Approximation Training data Galaxy Likelihood P h g Cool et al. (2007)
71 Population: FRANKEN-Z
72 How Valid is Our Approximation?
73 The Big Data Assumption P z g = P z h P h g, s g = 1, S g h h P z h P g h, b P h P g 1. What is this likelihood? 2. How do we compute it? 3. How do we deal with missing data? 4. How do we incorporate selection effects? 5. What is our prior?
74 The Big Data Assumption P z g = P z h P h g, s g = 1, S g h h P z h P g h, b P h P g 1. What is this likelihood? 2. How do we compute it? 3. How do we deal with missing data? 4. How do we incorporate selection effects? 5. What is our prior?
75 Missing Data How to deal with missing data? Training galaxy P g h, b Observed galaxy Band mask
76 Now What? How to deal with missing data? P g h, b 2 ln L =???
77 Naïve Likelihood: Multivariate Normal How to deal with missing data? P g h, b 2 ln L n χ n 2 δ n +n ln 2π + ln C g + C h
78 Implications How to deal with missing data? P g h, b SDSS Skyserver
79 Implications How to deal with missing data? P g h, b SDSS Skyserver SDSS Skyserver
80 Implications How to deal with missing data? P g h, b SDSS Skyserver SDSS Skyserver 2 ln L 5 χ ln 2π + ln C g + C h
81 Implications How to deal with missing data? P g h, b SDSS Skyserver SDSS Skyserver 2 ln L 3 χ ln 2π + ln C g + C h
82 Solutions How to deal with missing data? P g h, b SDSS Skyserver SDSS Skyserver 2 ln L X n X n, X n, X i.i.d. n χ 2 n δ n
83 Implementation How to deal with missing data? P g h, b 2 ln L =??? some transformation of doubly-non-central F- distribution
84 Implementation How to deal with missing data? P g h, b χ 2 = i ignore non-centrality 2 ln L = χ n 2 N b F g,i F h,i 2 σ 2 2 g,i + σ h,i first-order correction
85 Missing Data: Searching for Neighbors How to deal with missing data? P g h, b SDSS Skyserver Training data Galaxy Cool et al. (2007)
86 The Big Data Assumption P z g = P z h P h g, s g = 1, S g h h P z h P g h, b P h P g 1. What is this likelihood? 2. How do we compute it? 3. How do we deal with missing data? 4. How do we incorporate selection effects? 5. What is our prior?
87 The Big Data Assumption P z g = P z h P h g, s g = 1, S g h h P z h P g h, b P h P g 1. What is this likelihood? 2. How do we compute it? 3. How do we deal with missing data? 4. How do we incorporate selection effects? 5. What is our prior?
88 Selection Effects How to deal with selection effects? P g h
89 Selection Effects How to deal with selection effects? P F g g
90 Selection Effects Selection effect(s) How to deal with selection effects? P F g g, s g = 1, S g Binary selection flag (1=in/0=out)
91 Selection Effects Selection effect(s) How to deal with selection effects? P F g g, s g = 1, S g Selection Probability P s g = 1 F g, S g P F g g P s g = 1 g, S g Marginalized Selection Probability Original PDF Binary selection flag (1=in/0=out)
92 Application: Signal-to-Noise Cut
93 Application: Signal-to-Noise Cut
94 Application: Signal-to-Noise Cut
95 Application: Signal-to-Noise Cut
96 The Big Data Assumption P z g = P z h P h g, s g = 1, S g h h P z h P g h, b P h P g 1. What is this likelihood? 2. How do we compute it? 3. How do we deal with missing data? 4. How do we incorporate selection effects? 5. What is our prior?
97 The Big Data Assumption P z g = P z h P h g, s g = 1, S g h h P z h P g h, b P h P g 1. What is this likelihood? 2. How do we compute it? 3. How do we deal with missing data? 4. How do we incorporate selection effects? 5. What is our prior?
98 Mixing and Matching Galaxies How to deal with population mismatch? Based on Leistedt, Peiris, & Mortlock (2016)
99 Mixing and Matching Galaxies How to deal with population mismatch? p g ~ Mult n = 1, p = p g N h categories with probability vector p g = P h g Trinomial distribution from Mathworks. Based on Leistedt, Peiris, & Mortlock (2016)
100 Mixing and Matching Galaxies How to deal with population mismatch? p g ~ Mult n = 1, p = p g Trinomial distribution from Mathworks. N h categories with probability vector p g = P h g h = {0,1,2} p g = 0,0,1 h g = 2 Based on Leistedt, Peiris, & Mortlock (2016)
101 Mixing and Matching Galaxies How to deal with population mismatch? p g ~ Mult n = 1, p = p g P h n ~ p g g g Based on Leistedt, Peiris, & Mortlock (2016)
102 Mixing and Matching Galaxies How to deal with population mismatch? p g ~ Mult n = 1, p = p g n ~ p g g g P h Dir w α = n + 1 Population weights Concentration Counts Based on Leistedt, Peiris, & Mortlock (2016)
103 Mixing and Matching Galaxies How to deal with population mismatch? p g ~ Mult n = 1, p = p g Low concentration Courtesy of Wikipedia. n ~ High concentration p g g g P h Dir w α = n + 1 Population weights Concentration Counts Based on Leistedt, Peiris, & Mortlock (2016)
104 Hierarchical Modeling Hierarchical posteriors How to deal with population mismatch? P w, p g D, S Population weights (training set) Based on Leistedt, Peiris, & Mortlock (2016)
105 Hierarchical Modeling Hierarchical posteriors How to deal with population mismatch? P w, p g D, S Gibbs sampling P p g w, S Population weights (training set) P w p g, S Based on Leistedt, Peiris, & Mortlock (2016)
106 Hierarchical Modeling Hierarchical posteriors How to deal with population mismatch? P w, p g D, S Gibbs sampling P p g w, S Population weights (training set) 1. Sample hierarchical posteriors: i p g ~ Mult n = 1, p = p g w i 1 P w p g, S Based on Leistedt, Peiris, & Mortlock (2016)
107 Hierarchical Modeling Hierarchical posteriors How to deal with population mismatch? P w, p g D, S Gibbs sampling P p g w, S P w p g, S Population weights (training set) 1. Sample hierarchical posteriors: i p g ~ Mult n = 1, p = p g w i 1 2. Compute counts: n i ~ g g p g i Based on Leistedt, Peiris, & Mortlock (2016)
108 Hierarchical Modeling Hierarchical posteriors How to deal with population mismatch? P w, p g D, S Gibbs sampling P p g w, S P w p g, S Population weights (training set) 1. Sample hierarchical posteriors: i p g ~ Mult n = 1, p = p g w i 1 2. Compute counts: n i ~ g g p g i 3. Sample weights: w i Dir w n i + 1 Based on Leistedt, Peiris, & Mortlock (2016)
109 Application to Mock Data Errors ~ N, independent All objects share same PDF Counts from posterior (uniform prior) Counts from hierarchical model
110 Application to Mock Data Errors ~ N, independent All objects share same PDF Counts from posterior (uniform prior) Counts from hierarchical model
111 Application to Mock Data
112 Tests on ~100k SDSS Galaxies
113 Next Steps
114 Robustness of Training Data/Priors Hyper Suprime-Cam (HSC) SSP has ~380k objects taken from 11 surveys. Wide variety of selection criteria, data quality/reliability. Note: old plot from early internal data release.
115 Robustness of Training Data/Priors 1:1:1 prior 10:5:1 prior Spec-z : g/prism-z : photo-z Relative weights Preliminary
116 Incorporating Physical Priors Moving a galaxy from one redshift to another is a smooth, physical process that is well-understood. Want to incorporate this into our priors/predictions. GP with physical kernel Leistedt & Hogg (2016)
117 Incorporating Physical Priors Can use to augment training data. Straightforward to impute missing values. Leistedt & Hogg (2016)
118 Likelihood / Distance Metric What metric(s) to use? P g h Training galaxy More sophisticated machine learning methods could be used to compute posterior samples (or possibly likelihoods) over complex domains. Observed galaxy NGC 4414 NGC 5457
119 Summary Photometric redshifts (photo-z s) are an integral part of modern big data extragalactic science. Large training datasets gives new opportunities to develop Bayesian, data-driven photo-z s. Taking advantage of these datasets requires dealing with real-world problems (e.g., biased training data) using a variety of statistical methods (e.g., hierarchical Bayes). Early results look promising!
120 Code is available! Although still under active development, code, tutorials, and rough draft of a paper are online at: github.com/joshspeagle/frankenz
Inference of Galaxy Population Statistics Using Photometric Redshift Probability Distribution Functions
Inference of Galaxy Population Statistics Using Photometric Redshift Probability Distribution Functions Alex Malz Center for Cosmology and Particle Physics, New York University 7 June 2016 Galaxy population
More informationModern Image Processing Techniques in Astronomical Sky Surveys
Modern Image Processing Techniques in Astronomical Sky Surveys Items of the PhD thesis József Varga Astronomy MSc Eötvös Loránd University, Faculty of Science PhD School of Physics, Programme of Particle
More informationScience with large imaging surveys
Science with large imaging surveys Hiranya V. Peiris University College London Science from LSS surveys: A case study of SDSS quasars Boris Leistedt (UCL) with Daniel Mortlock (Imperial) Aurelien Benoit-Levy
More informationarxiv: v1 [astro-ph.co] 2 Dec 2016
Data-driven, interpretable photometric redshifts trained on heterogeneous and unrepresentative data Boris Leistedt 1, 2, 1, 3, 4, and David W. Hogg 1 Center for Cosmology and Particle Physics, Department
More informationProbabilistic photometric redshifts in the era of Petascale Astronomy
Probabilistic photometric redshifts in the era of Petascale Astronomy Matías Carrasco Kind NCSA/Department of Astronomy University of Illinois at Urbana-Champaign Tools for Astronomical Big Data March
More informationBayesian Modeling for Type Ia Supernova Data, Dust, and Distances
Bayesian Modeling for Type Ia Supernova Data, Dust, and Distances Kaisey Mandel Supernova Group Harvard-Smithsonian Center for Astrophysics ichasc Astrostatistics Seminar 17 Sept 213 1 Outline Brief Introduction
More informationHierarchical Bayesian Modeling
Hierarchical Bayesian Modeling Making scientific inferences about a population based on many individuals Angie Wolfgang NSF Postdoctoral Fellow, Penn State Astronomical Populations Once we discover an
More informationData Analysis Methods
Data Analysis Methods Successes, Opportunities, and Challenges Chad M. Schafer Department of Statistics & Data Science Carnegie Mellon University cschafer@cmu.edu The Astro/Data Science Community LSST
More informationLePhare Download Install Syntax Examples Acknowledgement Le Phare
LePhare Download Install Syntax Examples Acknowledgement http://www.lam.fr/lephare.html Le Phare A Photometric software to measure redshifts & galaxy properties Arnouts S. & Ilbert O. Basic concept SED
More informationLecture 11: SDSS Sources at Other Wavelengths: From X rays to radio. Astr 598: Astronomy with SDSS
Astr 598: Astronomy with SDSS Spring Quarter 4, University of Washington, Željko Ivezić Lecture : SDSS Sources at Other Wavelengths: From X rays to radio Large Surveys at Many Wavelengths SDSS: UV-IR five-band
More informationThe PRIsm MUlti-object Survey (PRIMUS)
The PRIsm MUlti-object Survey (PRIMUS) Alison Coil University of Arizona Steward Observatory March 2008 Overview: Galaxy evolution to z ~ 1 is still cosmic variance limited: DEEP2, VVDS, COMBO-17, COSMOS
More informationSED- dependent Galactic Extinction Prescription for Euclid and Future Cosmological Surveys
SED- dependent Galactic Extinction Prescription for Euclid and Future Cosmological Surveys Audrey Galametz MPE, Garching On behalf of the Euclid Photometric Redshift Organization Unit SED- dependent Galactic
More informationThe The largest assembly ESO high-redshift. Lidia Tasca & VUDS collaboration
The The largest assembly ESO high-redshift of massive Large galaxies Programme at 2
More informationSubaru/WFIRST Synergies for Cosmology and Galaxy Evolution
Artist s concept Subaru/WFIRST Synergies for Cosmology and Galaxy Evolution Dan Masters (JPL/California Institute of Technology) Collaborators: Peter Capak, Olivier Doré, Jason Rhodes, Shoubaneh Hemmati,
More informationDurham Lightcones: Synthetic galaxy survey catalogues from GALFORM. Jo Woodward, Alex Merson, Peder Norberg, Carlton Baugh, John Helly
Durham Lightcones: Synthetic galaxy survey catalogues from GALFORM Jo Woodward, Alex Merson, Peder Norberg, Carlton Baugh, John Helly Outline 1. How are the lightcone mocks constructed? 2. What mocks are
More informationDictionary Learning for photo-z estimation
Dictionary Learning for photo-z estimation Joana Frontera-Pons, Florent Sureau, Jérôme Bobin 5th September 2017 - Workshop Dictionary Learning on Manifolds MOTIVATION! Goal : Measure the radial positions
More informationHierarchical Bayesian Modeling
Hierarchical Bayesian Modeling Making scientific inferences about a population based on many individuals Angie Wolfgang NSF Postdoctoral Fellow, Penn State Astronomical Populations Once we discover an
More informationThe SDSS is Two Surveys
The SDSS is Two Surveys The Fuzzy Blob Survey The Squiggly Line Survey The Site The telescope 2.5 m mirror Digital Cameras 1.3 MegaPixels $150 4.3 Megapixels $850 100 GigaPixels $10,000,000 CCDs CCDs:
More informationWhat Can We Learn from Galaxy Clustering 1: Why Galaxy Clustering is Useful for AGN Clustering. Alison Coil UCSD
What Can We Learn from Galaxy Clustering 1: Why Galaxy Clustering is Useful for AGN Clustering Alison Coil UCSD Talk Outline 1. Brief review of what we know about galaxy clustering from observations 2.
More informationSNAP Photometric Redshifts and Simulations
SNAP Photometric Redshifts and Simulations Bahram Mobasher 1 Tomas Dahlen 2 1 University of California, Riverside 2 Space Telescope Science Institute Plan of the Talk Photometric Redshift technique- the
More informationA cooperative approach among methods for photometric redshifts estimation
A cooperative approach among methods for photometric redshifts estimation Stefano Cavuoti 1, Massimo Brescia 1, Giuseppe Longo 2, Valeria Amaro 2, Civita Vellucci 2 1 INAF - Observatory of Capodimonte,
More informationRefining Photometric Redshift Distributions with Cross-Correlations
Refining Photometric Redshift Distributions with Cross-Correlations Alexia Schulz Institute for Advanced Study Collaborators: Martin White Introduction Talk Overview Weak lensing tomography can improve
More informationExploiting Sparse Non-Linear Structure in Astronomical Data
Exploiting Sparse Non-Linear Structure in Astronomical Data Ann B. Lee Department of Statistics and Department of Machine Learning, Carnegie Mellon University Joint work with P. Freeman, C. Schafer, and
More informationADVANCED MACHINE LEARNING ADVANCED MACHINE LEARNING. Non-linear regression techniques Part - II
1 Non-linear regression techniques Part - II Regression Algorithms in this Course Support Vector Machine Relevance Vector Machine Support vector regression Boosting random projections Relevance vector
More informationAge = Gyr dependent on the detail physics.
Jiasheng Huang SAO Age = 13.3 13.7 Gyr dependent on the detail physics. Z=1 == D=10.8 x 10^9 light year, 5.9 Gyr since BB Z=3 == D=21.0 x 10^9 light year, 2.2 Gyr Z=5 == D=25.9 x 10^9 ly, 1.2 Gyr Z=7 ==
More informationSome Statistical Aspects of Photometric Redshift Estimation
Some Statistical Aspects of Photometric Redshift Estimation James Long November 9, 2016 1 / 32 Photometric Redshift Estimation Background Non Negative Least Squares Non Negative Matrix Factorization EAZY
More informationSelf-Organizing-Map and Deep-Learning! application to photometric redshift in HSC
Self-Organizing-Map and Deep-Learning! application to photometric redshift in HSC Atsushi J. Nishizawa (Nagoya IAR/GS of Sci.)! on behalf of HSC collaboration 2016 Nov. 25 @ Hiorshima University Today
More informationA Unified Theory of Photometric Redshifts. Tamás Budavári The Johns Hopkins University
A Unified Theory of Photometric Redshifts Tamás Budavári The Johns Hopkins University Haunting Questions Two classes of methods: empirical and template fitting Why are they so different? Anything fundamental?
More informationDark Energy. Cluster counts, weak lensing & Supernovae Ia all in one survey. Survey (DES)
Dark Energy Cluster counts, weak lensing & Supernovae Ia all in one survey Survey (DES) What is it? The DES Collaboration will build and use a wide field optical imager (DECam) to perform a wide area,
More informationSearching for z>6 Quasars with Subaru / Hyper Suprime-Cam Survey
Searching for z>6 Quasars with Subaru / Hyper Suprime-Cam Survey Masafusa Onoue (Sokendai / National Astronomical Observatory of Japan) on behalf of Subaru High-z Exploration of Low-Luminosity Quasars
More informationIntroduction. Chapter 1
Chapter 1 Introduction In this book we will be concerned with supervised learning, which is the problem of learning input-output mappings from empirical data (the training dataset). Depending on the characteristics
More informationData-driven models of stars
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
More informationHIERARCHICAL MODELING OF THE RESOLVE SURVEY
HIERARCHICAL MODELING OF THE RESOLVE SURVEY Sheila Kannappan, UNC Chapel Hill Faculty Fellow for SAMSI ASTRO Charlie Bonfield, UNC Chapel Hill Graduate Fellow for SAMSI ASTRO with input from Katie Eckert,
More informationTowards a Bayesian model for Cyber Security
Towards a Bayesian model for Cyber Security Mark Briers (mbriers@turing.ac.uk) Joint work with Henry Clausen and Prof. Niall Adams (Imperial College London) 27 September 2017 The Alan Turing Institute
More informationIN GALAXY GROUPS? WHERE IS THE CENTER OF MASS MATT GEORGE. 100 kpc 100 kpc UC BERKELEY
WHERE IS THE CENTER OF MASS IN GALAXY GROUPS? 100 kpc 100 kpc MATT GEORGE UC BERKELEY WITH ALEXIE LEAUTHAUD, KEVIN BUNDY, JEREMY TINKER, PETER CAPAK, ALEXIS FINOGUENOV, OLIVIER ILBERT, SIMONA MEI AND THE
More informationTreatment and analysis of data Applied statistics Lecture 4: Estimation
Treatment and analysis of data Applied statistics Lecture 4: Estimation Topics covered: Hierarchy of estimation methods Modelling of data The likelihood function The Maximum Likelihood Estimate (MLE) Confidence
More informationSurveys at z 1. Petchara Pattarakijwanich 20 February 2013
Surveys at z 1 Petchara Pattarakijwanich 20 February 2013 Outline Context & Motivation. Basics of Galaxy Survey. SDSS COMBO-17 DEEP2 COSMOS Scientific Results and Implications. Properties of z 1 galaxies.
More informationPhotometric Redshifts for the NSLS
Photometric Redshifts for the NSLS [ Northern Sky Legacy Survey ] S Arnouts (LAM) Main Goals To provide an optical survey to complement the Euclid NearIR + riz photometry in the Northern Sky. To design
More informationLyman-α Cosmology with BOSS Julián Bautista University of Utah. Rencontres du Vietnam Cosmology 2015
Lyman-α Cosmology with BOSS Julián Bautista University of Utah Rencontres du Vietnam Cosmology 2015 Lyman-α Forest of a Quasar (by Andrew Pontzen) F ( obs )= Observed flux Unabsorbed flux ( obs) 2 Lyman-α
More informationDusty star-forming galaxies at high redshift (part 5)
Dusty star-forming galaxies at high redshift (part 5) Flow of story 4.1.1 4.1.2 4.1.3 Millimetric Spectroscopic Redshifts Millimetric Photometric Redshifts Redshift Distributions of 24 μm selected DSFG
More informationQuasar identification with narrow-band cosmological surveys
Quasar identification with narrow-band cosmological surveys The case of J-PAS Silvia Bonoli and the J-PAS collaboration ESO, 28th June 2016 Hyper Suprime-Cam @ Subaru DESI Spectrocopic vs. - Provide SED
More informationNeutron inverse kinetics via Gaussian Processes
Neutron inverse kinetics via Gaussian Processes P. Picca Politecnico di Torino, Torino, Italy R. Furfaro University of Arizona, Tucson, Arizona Outline Introduction Review of inverse kinetics techniques
More informationSNLS supernovae photometric classification with machine learning
SNLS supernovae photometric classification with machine learning Courtesy of D. Campbell-Wilson Anais Möller CAASTRO, Australian National University Statistical Challenges in Astronomy, Carnegie Mellon
More informationHertzprung-Russel and colormagnitude. ASTR320 Wednesday January 31, 2018
Hertzprung-Russel and colormagnitude diagrams ASTR320 Wednesday January 31, 2018 H-R diagram vs. Color- Magnitude Diagram (CMD) H-R diagram: Plot of Luminosity vs. Temperature CMD: Plot of magnitude vs.
More informationMultistage Anslysis on Solar Spectral Analyses with Uncertainties in Atomic Physical Models
Multistage Anslysis on Solar Spectral Analyses with Uncertainties in Atomic Physical Models Xixi Yu Imperial College London xixi.yu16@imperial.ac.uk 23 October 2018 Xixi Yu (Imperial College London) Multistage
More information9/12/17. Types of learning. Modeling data. Supervised learning: Classification. Supervised learning: Regression. Unsupervised learning: Clustering
Types of learning Modeling data Supervised: we know input and targets Goal is to learn a model that, given input data, accurately predicts target data Unsupervised: we know the input only and want to make
More informationStatistical Tools and Techniques for Solar Astronomers
Statistical Tools and Techniques for Solar Astronomers Alexander W Blocker Nathan Stein SolarStat 2012 Outline Outline 1 Introduction & Objectives 2 Statistical issues with astronomical data 3 Example:
More informationIntroduction to Machine Learning Midterm Exam
10-701 Introduction to Machine Learning Midterm Exam Instructors: Eric Xing, Ziv Bar-Joseph 17 November, 2015 There are 11 questions, for a total of 100 points. This exam is open book, open notes, but
More informationStatistical Searches in Astrophysics and Cosmology
Statistical Searches in Astrophysics and Cosmology Ofer Lahav University College London, UK Abstract We illustrate some statistical challenges in Astrophysics and Cosmology, in particular noting the application
More informationThe magnitude system. ASTR320 Wednesday January 30, 2019
The magnitude system ASTR320 Wednesday January 30, 2019 What we measure: apparent brightness How bright a star appears to be in the sky depends on: How bright it actually is Luminosity and its distance
More informationCosmology on small scales: Emulating galaxy clustering and galaxy-galaxy lensing into the deeply nonlinear regime
Cosmology on small scales: Emulating galaxy clustering and galaxy-galaxy lensing into the deeply nonlinear regime Ben Wibking Department of Astronomy Ohio State University with Andres Salcedo, David Weinberg,
More informationLearning algorithms at the service of WISE survey
Katarzyna Ma lek 1,2,3, T. Krakowski 1, M. Bilicki 4,3, A. Pollo 1,5,3, A. Solarz 2,3, M. Krupa 5,3, A. Kurcz 5,3, W. Hellwing 6,3, J. Peacock 7, T. Jarrett 4 1 National Centre for Nuclear Research, ul.
More informationBayesian inference using Gaia data. Coryn Bailer-Jones Max Planck Institute for Astronomy, Heidelberg
Bayesian inference using Gaia data Coryn Bailer-Jones Max Planck Institute for Astronomy, Heidelberg What is Bayes? an approach to learning (= inference) given data on a phenomenon, determine how well
More informationGas Accretion & Outflows from Redshift z~1 Galaxies
Gas Accretion & Outflows from Redshift z~1 Galaxies David C. Koo Kate Rubin, Ben Weiner, Drew Phillips, Jason Prochaska, DEEP2, TKRS, & AEGIS Teams UCO/Lick Observatory, University of California, Santa
More informationImproved Astronomical Inferences via Nonparametric Density Estimation
Improved Astronomical Inferences via Nonparametric Density Estimation Chad M. Schafer, InCA Group www.incagroup.org Department of Statistics Carnegie Mellon University Work Supported by NSF, NASA-AISR
More informationarxiv:astro-ph/ v1 16 Apr 2004
AGN Physics with the Sloan Digital Sky Survey ASP Conference Series, Vol. 311, 2004 G.T. Richards and P.B. Hall, eds. Quasars and Narrow-Line Seyfert 1s: Trends and Selection Effects arxiv:astro-ph/0404334v1
More informationIntroduction to Machine Learning Midterm Exam Solutions
10-701 Introduction to Machine Learning Midterm Exam Solutions Instructors: Eric Xing, Ziv Bar-Joseph 17 November, 2015 There are 11 questions, for a total of 100 points. This exam is open book, open notes,
More informationSupplementary Information for SNLS-03D3bb a super- Chandrasekhar mass Type Ia supernova
1 Supplementary Information for SNLS-03D3bb a super- Chandrasekhar mass Type Ia supernova SN Location SNLS-03D3bb is located at RA: 14:16:18.920 Dec: +52:14:53.66 (J2000) in the D3 (extended Groth Strip)
More informationThree data analysis problems
Three data analysis problems Andreas Zezas University of Crete CfA Two types of problems: Fitting Source Classification Fitting: complex datasets Fitting: complex datasets Maragoudakis et al. in prep.
More informationSparse Composite Models of Galaxy Spectra (or Anything) using L1 Norm Regularization. Benjamin Weiner Steward Observatory University of Arizona
Sparse Composite Models of Galaxy Spectra (or Anything) using L1 Norm Regularization Benjamin Weiner Steward Observatory University of Arizona Inference of mass, redshift, etc. from physical models Many
More informationarxiv: v1 [astro-ph.ga] 14 May 2015
Publications of the Korean Astronomical Society pissn: 1225-1534 / eissn: 2287-6936 00: 1 99, 2014 March http://dx.doi.org/10.5303/pkas.2014.00.0.1 c 2014. The Korean Astronomical Society. All rights reserved.
More informationDust properties of galaxies at redshift z 5-6
Dust properties of galaxies at redshift z 5-6 Ivana Barisic 1, Supervisor: Dr. Peter L. Capak 2, and Co-supervisor: Dr. Andreas Faisst 2 1 Physics Department, University of Zagreb, Zagreb, Croatia 2 Infrared
More informationFrom quasars to dark energy Adventures with the clustering of luminous red galaxies
From quasars to dark energy Adventures with the clustering of luminous red galaxies Nikhil Padmanabhan 1 1 Lawrence Berkeley Labs 04-15-2008 / OSU CCAPP seminar N. Padmanabhan (LBL) Cosmology with LRGs
More informationAndy Casey Astrophysics; Statistics
Science in the era of Gaia data big Astrophysics; Statistics andycasey astrowizicist astrowizici.st Science in the era of Gaia data big - The Gaia mission All about Gaia. What makes data big? Science in
More informationLarge-scale Collaborative Prediction Using a Nonparametric Random Effects Model
Large-scale Collaborative Prediction Using a Nonparametric Random Effects Model Kai Yu Joint work with John Lafferty and Shenghuo Zhu NEC Laboratories America, Carnegie Mellon University First Prev Page
More informationChapter 7: From theory to observations
Chapter 7: From theory to observations Given the stellar mass and chemical composition of a ZAMS, the stellar modeling can, in principle, predict the evolution of the stellar bolometric luminosity, effective
More informationRecent Advances in Bayesian Inference Techniques
Recent Advances in Bayesian Inference Techniques Christopher M. Bishop Microsoft Research, Cambridge, U.K. research.microsoft.com/~cmbishop SIAM Conference on Data Mining, April 2004 Abstract Bayesian
More informationSTA 4273H: Statistical Machine Learning
STA 4273H: Statistical Machine Learning Russ Salakhutdinov Department of Computer Science! Department of Statistical Sciences! rsalakhu@cs.toronto.edu! h0p://www.cs.utoronto.ca/~rsalakhu/ Lecture 7 Approximate
More informationType Ia Supernovae: Standardizable Candles and Crayons
Type Ia Supernovae: Standardizable Candles and Crayons Ryan Foley Clay Fellow Harvard-Smithsonian Center for Astrophysics Collaborators: Stephane Blondin Dan Kasen Bob Kirshner Kaisey Mandel & Nathan Sanders
More informationKeck/Subaru Exchange Program Subaru Users Meeting January 20, 2011
Keck/Subaru Exchange Program Subaru Users Meeting January 20, 2011 Taft Armandroff, Director W. M. Keck Observatory With science results from: Drew Newman and Richard Ellis, Caltech A. Romanowsky, J. Strader,
More informationFormation and growth of galaxies in the young Universe: progress & challenges
Obergurgl. April 2014 Formation and growth of galaxies in the young Universe: progress & challenges Simon White Max Planck Institute for Astrophysics Ly α forest spectra and small-scale initial structure
More informationSubaru High-z Exploration of Low-Luminosity Quasars (SHELLQs)
Yoshiki Matsuoka (NAOJ) Subaru High-z Exploration of Low-Luminosity Quasars (SHELLQs) Yoshiki Matsuoka (NAOJ Ehime Univ.) on behalf of the SHELLQs collaboration SHELLQs Subaru High-z Exploration of Low-Luminosity
More informationEXONEST The Exoplanetary Explorer. Kevin H. Knuth and Ben Placek Department of Physics University at Albany (SUNY) Albany NY
EXONEST The Exoplanetary Explorer Kevin H. Knuth and Ben Placek Department of Physics University at Albany (SUNY) Albany NY Kepler Mission The Kepler mission, launched in 2009, aims to explore the structure
More informationThe ALMA z=4 Survey (AR4S)
The ALMA z=4 Survey (AR4S) ALMA studies of massive z~4-5 galaxies Corentin Schreiber Leiden University October 21 2016 In collaboration with: Maurilio Pannella, David Elbaz, Roger Leiton, Tao Wang, and
More information(Slides for Tue start here.)
(Slides for Tue start here.) Science with Large Samples 3:30-5:00, Tue Feb 20 Chairs: Knut Olsen & Melissa Graham Please sit roughly by science interest. Form small groups of 6-8. Assign a scribe. Multiple/blended
More information20x increase from z = 0 to 2!
black hole accretion (AGN) & star formation (SF) 20x increase from z = 0 to 2! due to more gas (initial supply or accretion) or higher efficiency gas è stars, AGN starbursts merging? ALMA survey of ISM
More informationApproximate Bayesian computation: an application to weak-lensing peak counts
STATISTICAL CHALLENGES IN MODERN ASTRONOMY VI Approximate Bayesian computation: an application to weak-lensing peak counts Chieh-An Lin & Martin Kilbinger SAp, CEA Saclay Carnegie Mellon University, Pittsburgh
More informationType II Supernovae as Standardized Candles
Type II Supernovae as Standardized Candles Mario Hamuy 1 2 Steward Observatory, The University of Arizona, Tucson, AZ 85721 Philip A. Pinto Steward Observatory, The University of Arizona, Tucson, AZ 85721
More informationIRS Spectroscopy of z~2 Galaxies
IRS Spectroscopy of z~2 Galaxies Houck et al., ApJ, 2005 Weedman et al., ApJ, 2005 Lutz et al., ApJ, 2005 Astronomy 671 Jason Marshall Opening the IR Wavelength Regime for Discovery One of the primary
More informationCosmological Merger Rates
Cosmological Merger Rates C. Brook, F. Governato, P. Jonsson Not Phil Hopkins Kelly Holley-Bockelmann Vanderbilt University and Fisk University k.holley@vanderbilt.edu Why do we care so much about the
More informationThe Long Faint Tail of the High-Redshift Galaxy Population
At the Edge of the Universe, October 2006, Sintra, Portugal ASP Conference Series, Vol. *, 2007 Eds. J. Afonso, H. Ferguson and R. Norris The Long Faint Tail of the High-Redshift Galaxy Population M. Sawicki
More informationMULTIPLE EXPOSURES IN LARGE SURVEYS
MULTIPLE EXPOSURES IN LARGE SURVEYS / Johns Hopkins University Big Data? Noisy Skewed Artifacts Big Data? Noisy Skewed Artifacts Serious Issues Significant fraction of catalogs is junk GALEX 50% PS1 3PI
More informationA Random Walk Through Astrometry
A Random Walk Through Astrometry Astrometry: The Second Oldest Profession George H. Kaplan Astronomical Applications Department Astrometry Department U.S. Naval Observatory Random Topics to be Covered
More informationTwo Statistical Problems in X-ray Astronomy
Two Statistical Problems in X-ray Astronomy Alexander W Blocker October 21, 2008 Outline 1 Introduction 2 Replacing stacking Problem Current method Model Further development 3 Time symmetry Problem Model
More informationDay 5: Generative models, structured classification
Day 5: Generative models, structured classification Introduction to Machine Learning Summer School June 18, 2018 - June 29, 2018, Chicago Instructor: Suriya Gunasekar, TTI Chicago 22 June 2018 Linear regression
More informationHierarchical Bayesian Modeling of Planet Populations
Hierarchical Bayesian Modeling of Planet Populations You ve found planets in your data (or not)...... now what?! Angie Wolfgang NSF Postdoctoral Fellow, Penn State Why Astrostats? This week: a sample of
More informationAstronomical image reduction using the Tractor
the Tractor DECaLS Fin Astronomical image reduction using the Tractor Dustin Lang McWilliams Postdoc Fellow Carnegie Mellon University visiting University of Waterloo UW / 2015-03-31 1 Astronomical image
More informationStages of a Big Project ***
Stages of a Big Project *** The Five stages of SDSS: 1. Denial 2. Anger 3. Bargaining 4. Depression 5. Acceptance *** With apologies to Elizabeth Kubler Ross The Carnegie Supernovae Project Wendy Freedman
More informationGaussian Processes for Audio Feature Extraction
Gaussian Processes for Audio Feature Extraction Dr. Richard E. Turner (ret26@cam.ac.uk) Computational and Biological Learning Lab Department of Engineering University of Cambridge Machine hearing pipeline
More informationThe Evolution of Massive Galaxies at 3 < z < 7 (The Hawaii 20 deg 2 Survey H2O)
D. Sanders, I. Szapudi, J. Barnes, K. Chambers, C. McPartland, A. Repp (Hawaii) P. Capak, I. Davidson (Caltech), S. Toft (Copenhagen), B. Mobasher (UCRiverside) The Evolution of Massive Galaxies at 3
More informationSource plane reconstruction of the giant gravitational arc in Abell 2667: a condidate Wolf-Rayet galaxy at z 1
Source plane reconstruction of the giant gravitational arc in Abell 2667: a condidate Wolf-Rayet galaxy at z 1 Speaker: Shuo Cao Department of Astronomy Beijing Normal University Collaborators: Giovanni
More informationBayesian Networks: Construction, Inference, Learning and Causal Interpretation. Volker Tresp Summer 2016
Bayesian Networks: Construction, Inference, Learning and Causal Interpretation Volker Tresp Summer 2016 1 Introduction So far we were mostly concerned with supervised learning: we predicted one or several
More informationRevealing new optically-emitting extragalactic Supernova Remnants
10 th Hellenic Astronomical Conference Ioannina, September 2011 Revealing new optically-emitting extragalactic Supernova Remnants Ioanna Leonidaki (NOA) Collaborators: P. Boumis (NOA), A. Zezas (UOC, CfA)
More informationClassification CE-717: Machine Learning Sharif University of Technology. M. Soleymani Fall 2012
Classification CE-717: Machine Learning Sharif University of Technology M. Soleymani Fall 2012 Topics Discriminant functions Logistic regression Perceptron Generative models Generative vs. discriminative
More informationReionization signatures in gamma-ray spectra
Reionization signatures in gamma-ray spectra Rudy Gilmore SISSA UC Santa Cruz Near-IR Background and the Epoch of Reionization AT&T Conference Center May 15, 2012 Outline Review: Modeling the evolving
More informationChapter 10: Unresolved Stellar Populations
Chapter 10: Unresolved Stellar Populations We now consider the case when individual stars are not resolved. So we need to use photometric and spectroscopic observations of integrated magnitudes, colors
More informationNaïve Bayes classification
Naïve Bayes classification 1 Probability theory Random variable: a variable whose possible values are numerical outcomes of a random phenomenon. Examples: A person s height, the outcome of a coin toss
More informationEUCLID Legacy with Spectroscopy
EUCLID Legacy with Spectroscopy Gianni Zamorani INAF - Bologna Astronomical Observatory (on behalf of the E-NIS Team) Observing the Dark Universe with Euclid 17-18 November 2009 ESTEC, The Netherlands
More informationLarge Scale Bayesian Inference
Large Scale Bayesian I in Cosmology Jens Jasche Garching, 11 September 2012 Introduction Cosmography 3D density and velocity fields Power-spectra, bi-spectra Dark Energy, Dark Matter, Gravity Cosmological
More informationThe evolution of the Galaxy Stellar Mass Function:
The evolution of the Galaxy Stellar Mass Function: towards the high-mass end at high redshifts Helena Domínguez Sánchez Jordi Cepa (IAC), Miguel Sánchez Portal (ESAC), Andrea Cimatti (UniBo), Francesca
More information