Bayesian Template-Based Approach to Classifying SDSS-II Supernovae from 3-Year Survey Brian Connolly Photometric Supernova ID Workshop 3/16/12.

Slides:



Advertisements
Similar presentations
Modeling of Data. Basic Bayes theorem Bayes theorem relates the conditional probabilities of two events A, and B: A might be a hypothesis and B might.
Advertisements

Dark energy workshop Copenhagen Aug Why the SNLS ? Questions to be addressed: -Can the intrinsic scatter in the Hubble diagram be further reduced?
CS479/679 Pattern Recognition Dr. George Bebis
Chapter 7 Title and Outline 1 7 Sampling Distributions and Point Estimation of Parameters 7-1 Point Estimation 7-2 Sampling Distributions and the Central.
Type Ia Supernovae in the SDSS Stripe 82 SDSS-KSG Workshop Du-Hwan Han and Changbom Park Kyungpook National University Kyungpook National.
Model Assessment, Selection and Averaging
Optimization of large-scale surveys to probe the DE David Parkinson University of Sussex Prospects and Principles for Probing the Problematic Propulsion.
Type Ia Supernovae in the Near-Infrared and the Ultraviolet Kevin Krisciunas Cook’s Branch Nature Conservancy, April 12, 2012.
SDSS-II SN survey: Constraining Dark Energy with intermediate- redshift probes Hubert Lampeitl University Portsmouth, ICG In collaboration with: H.J. Seo,
Galaxy and Mass Power Spectra Shaun Cole ICC, University of Durham Main Contributors: Ariel Sanchez (Cordoba) Steve Wilkins (Cambridge) Imperial College.
Eric Y. Hsiao University of Victoria.  less dust extinction  low redshift  standard candles in JHK bands  independent measure of Hubble constant 
Bayesian Decision Theory Chapter 2 (Duda et al.) – Sections
Star-Formation in Close Pairs Selected from the Sloan Digital Sky Survey Overview The effect of galaxy interactions on star formation has been investigated.
Use of neural networks for the identification of new z ≥ 3.6 radio QSOs from FIRST-SDSS DR5 R. Carballo Dpto. Matemática Aplicada y Ciencias de la Computación,
Predictive Automatic Relevance Determination by Expectation Propagation Yuan (Alan) Qi Thomas P. Minka Rosalind W. Picard Zoubin Ghahramani.
B12 Next Generation Supernova Surveys Marek Kowalski 1 and Bruno Leibundgut 2 1 Physikalisches Institut, Universität Bonn 2 European Southern Observatory.
1 SDSS-II Supernova Survey Josh Frieman Leopoldina Dark Energy Conference October 8, 2008 See also: poster by Hubert Lampeitl, talk by Bob Nichol.
Natalie RoeSNAP/SCP Journal Club “Identification of Type Ia Supernovae at Redshift 1.3 and Beyond with the Advanced Camera for Surveys on HST” Riess, Strolger,
SNLS : Spectroscopy of Supernovae with the VLT (status) Grégory Sainton LPNHE, CNRS/in2p3 University Paris VI & VII Paris, France On behalf of the SNLS.
Survey Science Group Workshop 박명구, 한두환 ( 경북대 )
Weidong Li Jesse Leaman Alex Filippenko Department of Astronomy University of California, Berkeley Nearby Supernova Rates from the Lick Observatory Supernova.
NAOKI YASUDA, MAMORU DOI (UTOKYO), AND TOMOKI MOROKUMA (NAOJ) SN Survey with HSC.
A Step towards Precise Cosmology from Type Ia Supernovae Wang Xiaofeng Tsinghua University IHEP, Beijing, 23/04, 2006.
Bayesian parameter estimation in cosmology with Population Monte Carlo By Darell Moodley (UKZN) Supervisor: Prof. K Moodley (UKZN) SKA Postgraduate conference,
● DES Galaxy Cluster Mock Catalogs – Local cluster luminosity function (LF), luminosity-mass, and number-mass relations (within R 200 virial region) from.
Padua: 1604 → 2004 – Supernovae as cosmological lighthouses SNLS – The SuperNova Legacy Survey Mark Sullivan (University of Toronto) on behalf of the SNLS.
SNLS: Overview and High-z Spectroscopy D. Andrew Howell (Toronto) for the SNLS Collaboration (see: for full list)
SN Ia Rates in the SNLS: Progress Report Mark Sullivan University of Oxford
Ensemble Classification Methods Rayid Ghani IR Seminar – 9/26/00.
1 SDSS Supernova Survey Josh Frieman Supernova Rates 2008, Florence May 19, 2008.
Pan-STARRS Photometric Classification of Supernovae using a Hierarchical Bayesian Model George Miller, Edo Berger, Nathan Sanders Harvard-Smithsonian Center.
Decelerating and Dustfree: Dark Energy Studies of Supernovae with the Hubble Space Telescope Kyle Dawson March 16, 2008 For the SuperNova Cosmology Project.
Statistics In HEP Helge VossHadron Collider Physics Summer School June 8-17, 2011― Statistics in HEP 1 How do we understand/interpret our measurements.
How Standard are Cosmological Standard Candles? Mathew Smith and Collaborators (UCT, ICG, Munich, LCOGT and SDSS-II) SKA Bursary Conference 02/12/2010.
Credit: O. Krause (Steward Obs.) et al., SSC, JPL, Caltech, NASA South Africa and the SDSS-II Supernova Survey Ed Elson (SAAO/NASSP) Bruce Bassett (SAAO/ICG)
Statistical Inference for the Mean Objectives: (Chapter 9, DeCoursey) -To understand the terms: Null Hypothesis, Rejection Region, and Type I and II errors.
Hunting youngest Type Ia SNe in the intermediate Palomar Transient Factory Yi Cao (Caltech) On behalf of the intermediate Palomar Transient Factory collaboration.
The SNLS has been allocated large amount of spectroscopic follow-up time at the VLT, Gemini North and South, Keck and Magellan. Example of a spectrum of.
Type Ia Supernovae and the Acceleration of the Universe: Results from the ESSENCE Supernova Survey Kevin Krisciunas, 5 April 2008.
SNAP Calibration Program Steps to Spectrophotometric Calibration The SNAP (Supernova / Acceleration Probe) mission’s primary science.
BAOs SDSS, DES, WFMOS teams (Bob Nichol, ICG Portsmouth)
Cosmic Inhomogeneities and Accelerating Expansion Ho Le Tuan Anh National University of Singapore PAQFT Nov 2008.
Pattern Classification All materials in these slides were taken from Pattern Classification (2nd ed) by R. O. Duda, P. E. Hart and D. G. Stork, John Wiley.
On Predictive Modeling for Claim Severity Paper in Spring 2005 CAS Forum Glenn Meyers ISO Innovative Analytics Predictive Modeling Seminar September 19,
Cosmic shear and intrinsic alignments Rachel Mandelbaum April 2, 2007 Collaborators: Christopher Hirata (IAS), Mustapha Ishak (UT Dallas), Uros Seljak.
Guest lecture: Feature Selection Alan Qi Dec 2, 2004.
SDSS II Supernova Survey - The Science Wednesday 29th August 2007 DARK Summer Institute Mathew Smith In collaboration with B. Nichol (ICG) and the SDSS.
Searching High-z Supernovae with HSC and WFMOS
NGC4603 Cepheids in NGC4603 Planetary Nebula Luminosity Function Number.
1 Baryon Acoustic Oscillations Prospects of Measuring Dark Energy Equation of State with LAMOST Xuelei Chen ( 陳學雷 ) National Astronomical Observatory of.
1 SDSS-II Supernova Survey Josh Frieman SDSS Science Symposium August 18, 2008.
Two useful methods for the supernova cosmologist: (1) Including CMB constraints by using the CMB shift parameters (2) A model-independent photometric redshift.
PHY306 1 Modern cosmology 2: Type Ia supernovae and Λ Distances at z ~1 Type Ia supernovae SNe Ia and cosmology Results from the Supernova Cosmology Project,
The Rate of Type Ia SNe at Redshift z=0.2 from SDSS-I Overlapping Fields Horesh Assaf, Dovi Poznanski, Eran Ofek, Prof. Dan Maoz SN Rates Florence.
Towards cosmology with a million supernovae: The BEAMS method Renée Hlozek Phys. Rev. D In collaboration with Bruce Bassett Martin Kunz.
In Bayesian theory, a test statistics can be defined by taking the ratio of the Bayes factors for the two hypotheses: The ratio measures the probability.
1.INTRODUCTION Supernovae Type Ia (SNeIa) Good candidate for standard candle to the high-z Universe (redshift
A Cosmology Independent Calibration of GRB Luminosity Relations and the Hubble Diagram Speaker: Speaker: Liang Nan Collaborators: Wei Ke Xiao, Yuan Liu,
Statistical Inference for the Mean Objectives: (Chapter 8&9, DeCoursey) -To understand the terms variance and standard error of a sample mean, Null Hypothesis,
Ch 1. Introduction Pattern Recognition and Machine Learning, C. M. Bishop, Updated by J.-H. Eom (2 nd round revision) Summarized by K.-I.
The Sloan Digital Sky Survey Supernova Search ● THANKS TO THE APO TEAM FROM THE SDSS-II SUPERNOVA SCIENCE TEAM!!
Lecture 1.31 Criteria for optimal reception of radio signals.
Reducing Photometric Redshift Uncertainties Through Galaxy Clustering
A photometric method to classify high-z supernovae found with HSC
Photometric redshift estimation.
Transfer Learning in Astronomy: A New Machine Learning Paradigm
CH 5: Multivariate Methods
Alan Qi Thomas P. Minka Rosalind W. Picard Zoubin Ghahramani
Photometric Redshift Training Sets
LSST Science: Supernova/Transients/variable stars
Presentation transcript:

Bayesian Template-Based Approach to Classifying SDSS-II Supernovae from 3-Year Survey Brian Connolly Photometric Supernova ID Workshop 3/16/12

Outline The Goal PSNID SDSS-II 3-year Supernova Survey Analysis Results Comments Conclusions

Goal “Use the 3 year SDSS-II survey as a test bed to identify photometric SN Ia candidates [using a Bayesian methodology] with realistic estimates of purity” (obtained by using spectroscopically confirmed Ia’s)

Plan Estimate purity and efficiency (FoM) with spectroscopic sample Type photometric sample assuming these estimates Results (e.g., see if we can do cosmology)

Motivation for Bayesian Approach  2 (Ia) simply deviation from Ia hypothesis Estimate p-values, tail probabilities, severely underestimate Type I error rate No alternative included (not true with other classical tests) No alternative exists in classical approach Gives you want you want (derived from logically consistent framework) Includes information about lesser fits vs Ronald Fischer Thomas Bayes

PSNID Simplest Template-Based Bayesian Classifier Directly sum over all templates for all parameters and all types to find

Evidences  2 includes uncertainties in the model (which give good  2 /dof for high S/N) Likelihoods

Priors Host-z: or flat Type Milky Way R V =3.1, SN R V =2.2 A V, T max,  Flat

Idea Kuznetsova and Connolly (2007) advocated using knowledge of P(Ia) in addition to best fit. Plot P(Ia) vs.  r, find region in spec confirmed sample that maximizes purity and efficiency in photometric sample

SDSS-II Three Year Supernova Survey Sept-Nov deg 2 region along celestial equator using 2.5m telescope ugriz 0.1<z<0.4 Cadence 4 days (average) >10,000 new variable and transients in differenced images Small number ID-ed as Ia’s PSNID and spec confirmed Largest uniform sample of SN candidates to date to study classification (3221 photometric candidates pass quality cuts, 2776 no spec observations) SPLIT SAMPLE INTO SPECTROSCOPICALLY CONFIRMED AND UNCONFIRMED

Templates Ia’s: Sako et al. (2008) CC: start with Nugent, Nugent et al. (2002), near SDSS light curves-II, D’Andrea et al. (2010) and choose those that maximize Ia purity (and efficiency) 24→8 CC templates

Spec. Confirmed Ia 2006jz, z=0.2 Marginalized A V and  distributions factor of 2 larger than w/ spec. redshift prior

Spec. Confirmed Ia 2005it, z=0.3 * Parameter estimation done with MCMC

Spectroscopically Confirmed Sample 508 SNIa 80 CC 202 AGN PSNID Analysis w/out Host Z PSNID Analysis w Host Z -5<t<5 one epoch >15 days S/N>5 in two gri bands One search season 367 SNIa 45 CC 83 AGN Candidates Small number of CC’s, account for this by comparing how galaxies targeted (have well defined selection criteria)

Cuts on Confirmed Sample (candidates near SDSS galaxy spectrum)

FoM (1) (2) (3) True Ia’s ID-ed as Ia’s True Ia’s after cuts (Contamination) Different for spec and photo sample

Results

Spectroscopically Unconfirmed Sample 3221 candidates 2776 no host-z 445 with host-z

Unconfirmed Sample 860 candidates 94% Purity 92% Efficiency  2 optimized for P(Ia)>0.9 Total Unconfirmed Sample

Toward a Hubble Diagram

Hubble Diagram

Comments

SN Challenge

pSNid II Template fitting Classification schemes –“Classical” –Color –Rising light curves –Sequential analysis Input: ascii, FITS, database Output: ascii

Software Package: pSNid II Does It All JPAS Study

Comments Reduced  2 taken with grain of salt Start cutting in parameter space giving up purely Bayesian framework (Kunz et al., loss functions, etc.) Can set purity and efficiency independently – SEQUENTIAL ANALYSIS METHODS Milton Friedman

Conclusions Described method of photometrically classifying a large SN sample with the help of a small spectroscopic subsample 1070 photometric SN Ia candidates from the SDSS-II SN Survey data 94% purity and 6% contamination Hubble diagram –Eliminating A V >1 eliminates problems M. Sako et al.m, The Astrophysical Journal, Volume 738, Issue 2, article id. 162 (2011).

backup

Tuning W to get Purity/Efficiency Correct in Photometric Sample Ia/CC ratio incorrect as Ia’s mag limited Choose a mag-limited sample of those SNe with galaxy spectra Take ratio of W=[P(Ia)>0.9]/[P(Ia)<0.1]