The Berkeley Transient Classification Pipeline project is a multi-departmental effort, combining people from the Astronomy, Computer Science, and Statistics departments as well as collaborators from Lawrence Berkeley Labs. Below is a list of publications relating to the pipeline.
- Near-infrared Observations of SN 2006gy
We present near-infrared (NIR) observations of the luminous SN 2006gy obtained two years following explosion with the 1.3-m robotic PAIRITEL telescope in Arizona and the NIRC2 LGS adaptive optics system on Keck. Our observations began less than a month after explosion and have continued for more than two years. We discuss how these NIR observations constrain the energetics of SN 2006gy and the possibility that this event was a pair instability supernova. AAM is supported by a NSF Graduate Research Fellowship.
- Rapid and Automated Classification of Events from the Palomar Transient Factory
The Transients Classification Pipeline (TCP) is a Berkeley-led project which federates data streams from multiple surveys and observatories, classifies with machine learning and astronomer-defined science priors, and broadcasts sources of interest to various science clients (using the VOEvent protocol). The TCP is a production-level project, designed for real-time analysis, being developed to handle the Palomar Transient Factory data stream. The TCP framework should scale to LSST data volumes.
Joshua S. Bloom, D. L. Starr, N. R. Butler, D. Poznanski, M. Rischard, R. Kennedy, J. Brewer
- The Standard Candle Method For Type II-P Supernovae - New Sample And PTF Perspective
Using a sample that is larger than used before, and yet more diverse, we show that Type II-plateau supernovae (SNe II-P) can be calibrated using the simple brightness to photospheric-velocity correlation presented by Hamuy & Pinto (2002), yielding a tight Hubble diagram, with a dispersion of 10% in distance, comparable to SNe Ia. We show that the descendant method of Nugent et al. (2006), might be further simplified as the correction for dust extinction has low statistical impact. This suggests that a distance measurement could be obtained with a single spectrum of a SN II-P combined with rest-frame I-band photometry at a similar time. If dust correction is applied, the best solution implies a very steep dust law, with RV < 2, as recently found for many Type Ia SNe. In addition, SNe 2003hl and 2003iq that both exploded in the same host galaxy NGC0772, give consistent distances, if a low value for RV is assumed.
D. Poznanski, P. Nugent, N. R. Butler, A. Gal-Yam, Joshua S. Bloom, M. Ganeshalingam, J. Silverman, A. Filippenko, W. Li, A. Miller, B. Cenko, N. Smith, PTF Collaboration
- A Web-Based Framework for Rapidly Building a Light Curve Warehouse
The Berkeley Transients Classification Pipeline (TCP) uses a machine-learning classifier to automatically categorize transients from large data torrents and provides automated notification of science events of interest. As part of the training process, we created a large warehouse of light-curve sources with well-labelled classes that serve as priors to the classification engine. This web-based interactive framework, which we are now making public (http://dotastro.org/), allows us to ingest light-curve data in a wide variety of formats and store it in a common internal data model. Data is passed between pipeline modules in a prototype XML representation of time-series format, which can also be emitted to collaborators through dotastro.org. After import, the sources can be visualized using Google Sky, light curves can be inspected interactively, and classifications can be manually adjusted.
- Real-time Source Classification using Berkeley's Transient Pipeline
The Berkeley Transients Classification Pipeline (TCP) is a parallelized source identification, classification, and broadcast pipeline which ingests several realtime data torrents and emits science events of pre-articulated interest. The TCP machine-learning algorithms are trained using a comprehensive science class hierarchy of light-curves which are resampled to emulate the cadence and quality of incoming observatory data streams. The referenced classified light-curves are contained within our ever-evolving public data warehouse (http://dotastro.org). To effectively distinguish a source's classification from neighboring classes or hierarchical parents, dozens of real-number metrics ("features") are derived from its light-curve, color info, spatial context, and historical data. Upon class identification (or reclassification), a VOEvent containing all available information is broadcast to subscribed telescopes and science groups for followup. Subsequently acquired data for that source can then be fed back to dotastro.org which the TCP will use to reinforce it's internal model of the source's science class.
Dan Starr, Josh Bloom, Maxime Rischard, Dovi Poznanski, John M. Brewer, Elizabeth Purdom, Chris Klein, Nat Butler