Aspects'of'Transient'Classifica2on!Ashish!Mahabal,!Caltech!C‐BAS,!UCI,!15!Feb!2011!Collaborators!• Caltech!– George'Djorgovski'– Ciro!Donalek!– Andrew!Drake!– Ma@hew!Graham!– Roy!Williams!• JPL!– Baback!Moghaddam!– Mike!Turmon!Plus!at!various!other!insMtutes!all!over,!but!especially!in!US,!India!and!Italy!2!Contact:[email protected][email protected]!Expanding!the!Observable!Parameter!Space!(M. Harwit)!Technology advances  Expanded domain of measurements  Discovery of new types of phenomena As we open up the time domain, we are bound to discover some new things!!SynopMc!surveys!Sporadic'to'repeated'SN!z=0.05!!CSS!20090711!!High‐amplitude!(non‐OT)!variables,!mainly!CVs!and!AGN,!over!the!Mme!baselines!~!a!few!years!!DPOSS'Plate'Overlap'Survey!DPOSS'Transients'(Mahabal,(Djorgovski,(Grane5(2001,(2003)(The!Palomar‐Quest!(PQ)!Digital!SynopMc!Sky!Survey!• Palomar!48‐in.!+!112‐CCD,!161!Mpix!camera!• A!Caltech‐Yale!collab.!Co‐PIs:!C.!Baltay!&!!SGD;!plus!other!groups!worldwide!(LBL,!etc.)!• Many!passes!with!up!to!4!filters!(UBRI/griz),!Mme!baselines!from!minutes!to!years!• Collected!>!50!TB!of!data!• Operated!from!Aug.!2003!through!Sept.!2008!• Key$goal:$$Explora.on$of$the$.me$domain$h@p://palquest.org' !!LBL!SNF!search!(Nugent!et!al.)!>!700!SNe!discovered!The!Most!Variable!Sources!on!the!Sky:!PQV 2325-0140!PQV 2321-1448!Cataclysmic!Variables!and!Dwarf!Novae!Blazars!and!OVV!Quasars!Selected(in(the(Palomar‐Quest(Survey(Catalina!Sky!Survey(s):!CRTS uses the data from all three Catalina NEO surveys, with a coverage of up to 2,500 deg2 / night, and the total area coverage of ~ 30,000 deg2 We are processing the Catalina data streams in real time to look for astrophysical transients NEO survey Co-PI’s: E. Beshore & S. Larson (LPL)CSS!Discoveries!of!Earth‐Grazing!Asteroids!November 6, 2009 CSS has discovered ~70% of all known NEOs An extremely low cost “sample return mission”PQ!CSS!LSST!Astrosat!PTF!Skymapper!Pan‐STARRS!Recent,'current'and'future'mul2epoch'surveys'Orders'of'magnitudes'different.'Move'towards'digital'movies!'GALEX,!Spitzer,!FIRST,!…!The!Palomar‐Quest!Event!Factory!R I tonight baseline Detect ~ 1 - 2 106 sources per half-night scan Find ~ 103 apparent transients (in the data) Identify ~ 2 - 4 102 real transients (on the sky) Identify ~ 1 - 10 possible Astrophysical transients Compare with the baseline sky Remove instrum. artifacts Remove asteroids Classification and follow-up Sept. 2006The!Tsunami!Wave!of!the!Future!• Forthcoming!on!a!Mme!scale!~!1!‐!5!years:!~'1'TB'/'night,'~104'transients'/'night!(PanSTARRS,!Skymapper,!VISTA,!VST…)!• Forthcoming!in!~!8!‐!10!years:!LSST,!~'30'TB'/'night,'~'105''transients'/'night'• ObservaMonal!follow‐up!needs:!– Rapid!photometric/posiMonal!monitoring!– Rapid!spectroscopy!– InformaMon/computaMon!infrastructure!A major, qualitative change! • Now: data streams of ~ 0.1 TB / night, ~ 10 - 102 transients / night (CRTS, PQ, PTF, various SN surveys, asteroid surveys) } } Transient classification technologies are essentialTime'Domain'Astrophysics!• Moving'objects:!!Solar!system,!GalacMc!structure,!exoplanets!• Variability!Physical'causes'of'intrinsic'variability:!– EvoluMon!(structural!changes!etc.),!generally!long!Mme!scales!– Internal!processes,!e.g.,!turbulence!inside!stars!– AccreMon!/!collapse,!protostars!to!CVs!to!GRBs!to!QSOs!– Thermonuclear!explosions!– MagneMc!field!reconnecMons,!e.g.,!stellar!flares!– Line!of!sight!changes!(rotaMon,!jet!wig gles…)!Variability'is'known'on'2me'scales'from'ms'to'1010'yr'Synop2c,'panoramic'surveys''event'discovery'Rapid'follow‐up'and'mul2‐λ''keys'to'understanding'Intrinsic!Modulation along the LOS: microlensing, ISS,! eclipses, variable extinction …!Intrinsically'Variable'Phenomena!• Things!we!know!about:!– Stars:!!oscillaMons,!noise,!acMvity!cycles,!atmospheric!phenomena!(flares,!etc.),!eclipses,!explosions!(SNe,!GRBs),!accreMon!(CVs,!novae),!spinning!beams!(pulsars,!SS!433,!…)!– AGN:!accreMon!power!spectrum,!beaming!phenomena!• Things!we!see,!but!don’t!really!understand:!– Faint!fast!transients!– Archival!OTs!– Megaflares!on!normal!stars!• Things!we!expect!to!see,!and!maybe!we!do:!– Breakout!shocks!of!Type!II!SNe!– SMBH!loss!cone!accreMon!events!– BH!mergers!(LIGO,!LISA?),!QSO!formaMon…?!Examples!of!CRTS!Transients!CSS090429:135125‐075714!Flare!star!CSS090429:101546+033311!Dwarf!Nova!CSS090426:074240+544425!Blazar,!2EG!J0744+5438!!Vastly!different!physical!phenomena,!and!yet!they!look!the!same!Which!ones!are!the!most!interesMng!and!worthy!of!follow‐up?!Rapid,$automated$transient$classifica.on$is$a$cri.cal$need!$Semantic Tree of Astronomical Variables and Transients AGN!Subtypes!SN!Subtypes!Towards!Automated!Event!ClassificaMon!Event'parameters:'m1(t),!m2(t),…!α,!δ,!µ,!…!image!shape…!colors! lightcurves!etc.!Expert!and!ML!generated!priors!contextual!informaMon!Event'Classifica2on'Engine'P(SN!Ia)!=!…!P(SN!II)!=!…!P(AGN)!=!…!P(CV)!=!…!P(dM)!=!…!….!A!necessity!for!large!synopMc!surveys!ClassificaMon!probabiliMes!(evolving,!iterated)!IVOA'With!M!Turmon!and!B!Moghaddam,!JPL!17!What!is!a!transient?!Fast!transient!(flaring!dM),!CSS080118:112149–131310!4!individual!exposures,!separated!by!10!min! Light!curve!Something'that'has'a'large'delta‐magnitude'for'a'small'delta‐2me'Sample!Light!Curves!Blazar!PKS0823+033!Variables!and!transients!–!the!disMncMon!is!one!of!percepMon,!and!your!aims!CV!111545+425822!Supernova!The!importance!of!context!18!June!2010! AstroinformaMcs!2010!Broad,!incomplete!hierarchy!All!transients!SN!SN!I!CV!CV,!blazars,!periodic!SN!II!CV,!blazars!periodic!blazars!To'other'classifiers'• More!context!informaMon!will!help!• How!to!maintain!uniformity?!• What!when!filters!change?!• Uniformity!of!priors?!• Number!of!objects!• Their!magnitude!range!• Spread!over!Mme!• Ground!truth?!QuesMons!raised!by!the!!Data!paucity!regime!• How!many!classes?!• Too!few:!probabiliMes!incorrect!(where!do!objects!belonging!to!unrepresented!classes!go?)!• Too!many:!overlaps!increase!(e.g.!SN!of!different!types;!variables!of!different!types)!and!probability!splits!into!smaller!fracMons!•