Data Challenges

In order to encourage development of gravitational wave (GW) search algorithms for pulsar timing data, the IPTA has issued the first “IPTA Data Challenge”, i.e., simulated pulsar data containing an unknown GW signal. The challenge is to develop algorithms to detect, or limit the presence of, a GW signal in the data. This, and all future data challenges are open to all, and submissions from external parties are encouraged.

IPTA data challenges will consist of “open” and “closed” data sets. The answer (i.e. the full detail of the simulated signals) for the open data set will be published at the start of the data challenge and can be used for calibration and tests of software. The closed data sets are the real challenge, for which all information regarding the simulations will remain confidential until after the close of the data challenge.

The data challenges are produced by the IPTA data challenge committee, consisting of one member from each of the IPTA partner collaborations. For more information, please contact one of the IPTA data challenge committee members:

NANOGrav:    Fredrick A. Jenet         (fredrickajenet at gmail.com)
PPTA:              Michael Keith               (mkeith at pulsarastronomy.net)
EPTA:              KJ Lee                            (kjlee at mpifr-bonn.mpg.de)

Data Challenge 1

Data challenge issued: 2012-03-23; Data challenge closes: 2012-09-28

The First IPTA Data Challenge was released on 23rd March 2012 and will close on 28th September 2012, with results released shortly afterwards. This first challenge simplifies some of the complexities of pulsar timing array data, and is therefore a “warm up” to the more difficult data challenges that follow. For more information, or to download the data sets, go to the Data Challenge 1 page.

Parameters used for the first closed data challenge are now available here

Data Challenge 1 results

The figure above summarizes the data challenge 1 results submitted to the challenge committee. Algorithms assuming a spectral index are shown in the top three panels, which algorithms which search for a spectral index are shown in the bottom panel.

In general we found that all algorithms recover the correct parameters for the simplified data sets of challenge one.


Comments are closed.