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 projects using data products and tools developed for the TCP project.
- SimpleTimeseries
The enormous flood of data expected from synoptic surveys in coming years has the opportunity to open up countless new science opportunities. However, without efficient ways to share that data the science may sit untapped and undiscovered. SimpleTimeseries is an XML data format designed for the TCP project and generalized with the help of the VOEvent Working Group. With a standard way to share time-series information, we can concentrate on novel discoveries instead of data-mining. More information can be found on the SimpleTimeseries pages here on dotastro.org.
John Brewer, Josh Bloom, Dan Starr, and the VOEvent Working Group
- AM CVn
AM CVn stars are an underrepresented subclass of Cataclysmic Variable stars, characterized by Helium-dominated spectra and orbital periods of 10 -- 60 minutes. Using the wealth of photometric data collected during the SDSS Stripe 82 survey, we are hoping to identify low-amplitude variable sources with colors similar to known AM CVn stars as strong new AM CVn candidates. The lightcurve analysis techniques developed will also be useful for identifying variable sources in future large-scale photometric surveys.
Joshua Shiode
- TCP on CITRUS
Although the current TCP handles classifying an incoming data stream nearing several dozen sources per second, future telescope surveys will have data rates several hundred times larger and a new architecture will be needed. We are interested in exploring Hadoop based classification algorithms, such as Apache Mahout or similar, as an alternative method for time series classification with these next generation telescopes. You can find more information at the CITRUS site.
- Core Collapse Supernovae
Dovi Poznanski is currently working on the study of core collapse supernovae, and on using supernovae II-P for cosmography. Both require large well defined samples, a natural product of applying TCP to upcoming surveys.
- PyMPChecker
A Python based rogram and webservice to predict asteroid locations for query coordinates. It is now available to the community through a website offering individual and batch query services. It will help TCP by quickly determining if a transient is likely to be an asteroid or not.
Chris Klein, Dan Starr and Gareth Williams at the Minor Planet Center.
- Noisification
The machine-learning algorithms need to be trained on real-world, modeled, and nosified data. This is a project to model the light curves of well-observed transients and variables and then nosify the data to produce many times more input light curves to help teach the classification algorithm.
- High Proper Motion
To provide additional training and testing data, Josh Bloom and Dan Star used a custom pipeline to create a catalog of sources from SDSS Stripe 82 FITS images. John Brewer is now deriving proper motion and parallax data for those sources and using the catalog to search for interesting high proper motion objects. Proper motion and parallax data will also be used to enhance classification of sources in the TCP.
- TCP Tutor
The classifications of initial training data for the classification enging need to be well known. Decades of well studied and labled data exists in public archives but the wide variety of data formats and labelling schemes make them difficult to use. The TCP Tutor project defined a common data format and built a set of tools to quickly import this heterogeneous data as well as visualize the lightcurves and locations on the sky. Additionally, it can emit that data in a well defined XML format for use by the other parts of the TCP system.