A Gaussian Process Surrogate Model Assisted Multi-optimization Algorithm for Pulsar Period Searching

In recent years, X-ray pulsar-based navigation and timing have been widely concerned. Period searching is a key technique of pulsar navigation and pulsar timing. Since the pulsar signal is extremely weak, an accurate result of pulsar period requires a large amount of pulsar photons. However, the current method for pulsar period searching is of high computational cost when the amount of pulsar photons is very large. In this paper, we propose a Gaussian process (GP) surrogate model assisted multi-optimization algorithm to reduce the computational cost of period searching. In this algorithm, GP with low computational cost is used as a surrogate of the objective function of period searching. Besides, the proposed multi-optimization algorithm combines the advantages of Particle swarm optimization (PSO) and Cross-Entropy (CE) algorithm, which guarantees the accuracy of algorithm. The performance of the proposed algorithm is verified by the experiments with the simulation dada of PSR B1821-24 and Crab pulsar. Experiment results shows that the proposed method can efficiently reduce the computation cost while remains the accuracy of period searching. Besides, the proposed algorithm has good universality that can be used in different period searching method such as the epoch folding method and the MLE method.