Nonlinear time series prediction based on a power-law noise model

In this paper we investigate the influence of a power-law noise model, also called noise, on the performance of a feed-forward neural network used to predict time series. We introduce an optimization procedure that optimizes the parameters the neural networks by maximizing the likelihood function based on the power-law model. We show that our optimization procedure minimizes the mean squared leading to an optimal prediction. Further, we present numerical results applying method to time series from the logistic map and the annual number of sunspots demonstrate that a power-law noise model gives better results than a Gaussian model.

The POINT-AGAPE survey: 4 high signal-to-noise microlensing candidates detected towards M 31

We have carried out a survey of the Andromeda galaxy for unresolved microlensing (pixel lensing). We present a subset of four short timescale, high signal-to-noise microlensing candidates found by imposing severe selection criteria: the source flux variation exceeds the flux of an R = 21 magnitude star and the full width at half maximum timescale is less than 25 days. Remarkably, in three out of four cases, we have been able to measure or strongly constrain the Einstein crossing time of the event. One event, which lies projected on the M 31 bulge, is almost certainly due to a stellar lens in the bulge of M 31. The other three candidates can be explained either by stars in M 31 and M 32 or by MACHOs.