|Description:|| The LOFAR Transients Pipeline (TraP; Swinbank et al. 2015) has become a highly effective tool for processing many thousands of radio images from a number of different radio telescopes. However, it can be challenging to find all of the transient and variable sources lurking in the TraP database. Using simulated transients and a large number of snapshot observations from the LOFAR Radio Sky Monitor (Fender et al. 2008), we have developed machine learning algorithms to classify our sources (Rowlinson et al. 2019, accepted by Astronomy & Computing).|
This image shows the results from most successful method, a simple logistic regression algorithm. We plot two variability parameters from the TraP database. The grey region illustrates all of the sources from the real snapshot observations. The blue data points are the true positive transient detections (correctly found simulated transients) and the green data points are those simulated sources that were not found (too faint or insufficiently variable). The red data points are sources from the real sources that the algorithm has labelled as transients, two of which are known scintillating pulsars.
For more info, the paper is available here: http://adsabs.harvard.edu/abs/2018arXiv180807781R