iPTF Discoveries of Recent Core-Collapse Supernovae
ATel #7112; F. Taddia, S. Papadogiannakis, J. Johansson, R. Ferretti, C. Fremling, K. Migotto, A. Nyholm, T. Petrushevska, R. Roy, E. Karamehmetoglu (OKC), G. Hosseinzadeh (UCSB/LCOGT), S. Ben-Ami, A. De Cia, Y. Dzigan, A. Horesh, G. Leloudas, I. Manulis, A. Rubin, I. Sagiv, P. Vreeswijk, O. Yaron (Weizmann), Y. Cao, G. Duggan, S. Kulkarni, D. Perley, P. Bilgi, (Caltech)
on 19 Feb 2015; 17:32 UT
Distributed as an Instant Email Notice Supernovae
Credential Certification: Francesco Taddia (ftadd@astro.su.se)
Subjects: Optical, Supernovae, Transient
The intermediate Palomar Transient Factory (ATel #4807)
reports the discovery and classification of the following Core-Collapse SNe. The approved classification (and/or subsequent) spectra are made publicly available through WISeREP (Yaron & Gal-Yam 2012).
Name | RA (J2000) | Dec (J2000) | Discovery date | Mag. | Abs. Mag. | Band | Redshift | Type | Spec. date | Instrument | Observers/Reducers | Scanner | RB2/RB4/RB5 |
iPTF15ak | 03:29:14.25 | +14:50:52.9 | 2015-01-16.16 | 20.6 | -18.0 | PTF-g | 0.11 | SN IIn | 2015-01-22 | Keck1+LRIS | G. Duggan, S. Kulkarni / D. Perley | E. Karamehmetoglu | 0.79/0.45/0.93 |
iPTF15jg | 12:27:19.15 | +07:17:02.3 | 2015-02-12.41 | 19.7 | -15.4 | PTF-g | 0.0251 | SN II | 2015-02-13 | P200+DBSP | G. Duggan, Y. Cao , P. Bilgi, S. Papadogiannakis / P. Vreeswijk | A. Nyholm | 0.47/0.57/0.95 |
[Dates are in UT.]
Transients are identified using a two-stage procedure.
High probability candidates are first filtered using a fully automated system that
employs machine learning algorithms. This sample is then vetted by human
"scanners" who produce a shorter, prioritized list of the most promising events
for follow-up. The legacy machine learning classification developed by
PTF is based on randomized collections of decision trees known as random forests (RB2; Brink et al. 2013MNRAS.435.1047B).
The new system introduced in iPTF extends this successful approach in several directions to boost performance while providing a test bed for new ideas.
It utilizes sparse dictionary representations to automatically extract informative features directly from pixel data (RB5; Wozniak et al. 2013AAS...22143105W)
and combines larger, better quality training samples with a feature space redesigned to address the data shift between PTF and iPTF (RB4; Rebbapragada et al. 2015AAS...22543402R).
With the current combined system it is possible to reject 99% of false alarms while missing less than 6% of real transients.
Future development will focus on early change detection, handling multiple photometric bands and online learning from incremental data.