Supernovae photometric classification of SNLS data with supervised learning

Transcription

1 Supernovae photometric classification of SNLS data with supervised learning Anais Möller CAASTRO & Australian National University A. Möller,V. Ruhlmann-Kleider, C. Leloup,J. Neveu, N. Palanque-Delabrouille,J. Rich, R. Carlberg,C. Lidman, C. Pritchet. (2016 in prep.)

2 Type Ia SNe very luminous stellar explosions can be used as standard candles

3 Type Ia SNe 1.0 very luminous stellar explosions time varying? can be used as standard candles expansion of the Universe 0.5 Betoule et al wa Betoule et al Planck+WP+JLA Planck+WP+C11 Planck+WP+BAO+JLA Planck+WP+BAO w 0 w(a) = w 0 + w a (1 a) with a = 1/(1 + z)

4 SNIa cosmology challenges we want : - more SNe Ia - precise measurements e.g. improving systematics. e.g. JLA Betoule et al host galaxy - modelling / templates - anchoring the HD -.

5 SNe pipeline in a nutshell transient classified photometry sample spectroscopy SN Ia sample redshift not enough resources for spectroscopy (DES,LSST) -> photometric classification

6 SNe pipeline in a nutshell transient classified photometry sample spectroscopy SN Ia sample photometry redshift

7 find the SNe Ia (or find the small yellow fish ) you need: - fish features: is it long, small, color? - a decision

8 SN photometric classification challenges we need: 1. light curve features 2. redshift 3. a classification strategy

9 SN photometric classification challenges we need: 1. light curve features 2. redshift 3. a classification strategy we want: 1. to trust our sample (purity) 2. large number (efficiency)

10 SN photometric classification challenges evaluate: we need: 1. light curve features 2. redshift 3. a classification strategy simulation e.g. Supernova photometric classification challenge Kessler et al N. Karpenka et al M. Lochner et al we want: 1. to trust our sample (purity) 2. large number (efficiency) data need a large known SN sample to test G. Bazin 2010

11 SNLS the SuperNova Legacy Survey based on the Canada France Hawaii Telescope MegaCam : 36 CCD mosaic 4 broadband filters g,r,i,z 4 fields of 1 square degree rolling search mode spectroscopic follow up (Keck, Gemini and VLT) observations: SNLS3 analysed and published - SNLS5 currently being processed (complete SNLS data set)

12 SNLS deferred photometric pipeline developed in the SNLS Saclay group (France) SN-like sample classified SN Ia sample (looks like a fish)

13 SNLS deferred photometric pipeline G. Bazin et al. A&A 534, A43 (2011) SN-like sample classified SN Ia sample Selection cuts - quality cuts on fit - color sampling Classification 1. redshift: host-galaxy photometric 2. light-curve fitter: SALT2 3. strategy: sequential cuts SNLS3 data spectroscopic SNe Ia spectroscopic CC simulated SNe Ia

14 SNLS deferred photometric pipeline G. Bazin et al. A&A 534, A43 (2011) SN-like sample classified SN Ia sample Selection cuts - quality cuts on fit - color sampling Classification 1. redshift: host-galaxy photometric 2. light-curve fitter: SALT2 3. strategy: sequential cuts P = 94.4 ± 0.5% 278 spe. SNe Ia 486 photometrically classified type Ia!

15 new photometric classification

16 SNLS new photometric classification pipeline SN-like sample classified SN Ia sample Selection cuts Classification 1. redshift: SN photometric z 2. general light-curve fitter 3. supervised learning simulations data SNLS3

17 SNLS new photometric classification pipeline SN-like sample classified SN Ia sample Selection cuts data SNLS3 found to be very important non SN-like backgrounds! all events spectroscopic SNe Ia spectroscopic CC visually non SN-like visually long extreme z

18 SNLS new photometric classification pipeline SN-like sample classified SN Ia sample Selection cuts Classification 1. redshift: SN photometric z Palanque-Delabrouille et al estimated directly from SN light curves. iterative: SALT2 fitted with different z, priors ~2% resolution catastrophic assignment ( z/(1 + z) > 0.15 ): 1.4%

19 SNLS new photometric classification pipeline SN-like sample classified SN Ia sample Selection cuts Classification 1. redshift: SN photometric z Palanque-Delabrouille et al estimated directly from SN light curves. iterative: SALT2 fitted with different z, priors ~2% resolution catastrophic assignment ( z/(1 + z) > 0.15 ): 1.4% simulations

20 SNLS new photometric classification pipeline SN-like sample classified SN Ia sample Selection cuts Classification 1. redshift: SN photometric z 2. general light-curve fitter independent from SALT2 fits other SNe f k = A k e (t tk 0 )/ k fall 1+e (t tk 0 )/ k rise k= filter t k max = t k 0 + k riseln( k fall/ k rise 1) + c k

21 SNLS new photometric classification pipeline SN-like sample classified SN Ia sample Selection cuts Classification 1. redshift: SN photometric z 2. general light-curve fitter 3. supervised learning in a nutshell supervised learning train simulations evaluate simulations data SNLS3

22 SNLS new photometric classification pipeline SN-like sample classified SN Ia sample Selection cuts Classification 1. redshift: SN photometric z 2. general light-curve fitter 3. supervised learning in a nutshell Decision Trees: - Random Forest - AdaBoost - XGBoost scikit-learn

23 SNLS new photometric classification pipeline SN-like sample classified SN Ia sample Selection cuts Classification 1. redshift: SN photometric z 2. general light-curve fitter 3. supervised learning

24 new photometric classification performance of our classifiers very good! simulations

25 new photometric classification if we want to do cosmology with a photometrically classified type Ia sample we must answer some important questions

26 new photometric classification differences between simulation & data? case study: purity 95% AdaBoost Random Forest XGBoost total e ciency Ia 36.9 ± ± ± 0.7 purity Ia 95.6 ± ± ± 0.4 contamination Ia inaccurate z 0.53 ± ± ± 0.09 contamination CC 3.9 ± ± ± 0.4 simulations AdaBoost Random Forest XGBoost photometric sample spectroscopic Ia photometric Ia spectroscopic CC photometric CC data SNLS3

27 new photometric classification is my sample eff/purity changing with z? simulations

28 new photometric classification what is the best we can do? XGBoost purity 98% cut SNLS3 events spectroscopic photometric simulated in sample Ia CC Ia CC Ia% SN-like selected classified data SNLS3

29 new photometric classification interesting things all events classified events (a) synthetic SNe in our photometric sample - light-curve quality and classification - spectroscopic confidence index and photometric classification - what about peculiar type Ia? and subluminous?

30 summary new photometric classification with supervised learning: we want: 1. trust our sample (purity) 2. large number (efficiency) new algorithm SN photometric z supervised classification evaluated with: simulations and data method to be applied to SNLS5 remember: supervised learning relies on training. found selection cuts very important. cosmology & photo SNe Ia? A. Möller,V. Ruhlmann-Kleider, C. Leloup,J. Neveu, N. Palanque-Delabrouille,J. Rich, R. Carlberg,C. Lidman, C. Pritchet. (2016 in prep.)