Multi-Objective Optimisation in Time Series: Time Delay Agreement

摘要

Several time delay estimates have been reported for the quasar Q0957+561. They come from distinct data sets and published separately. This paper presents a methodology to estimate a single time delay given several data sets by using multi-objective optimisation. We use General Regression Neural Networks (GRNN) to estimate the time delay, which is one of the most accurate time delay estimators - and faster. For the time delay agreement, we use hill-climbing search. We found that the best agreement for the time delay on Q0957+561 is Δ = 420 days.