MaxEnt power spectrum estimation using the Fourier transform for irregularly sampled data applied to a record of stellar luminosity

The principle of maximum entropy is applied to the spectral analysis of a data signal with general variance matrix and containing gaps in the record. The role of the entropic regularizer is to prevent one from overestimating structure in the spectrum when faced with imperfect data. Several arguments are presented suggesting that the arbitrary prefactor should not be introduced to the entropy term. The introduction of that factor is not required when a continuous Poisson distribution is used for the amplitude coefficients. We compare the formalism for when the variance of the data is known explicitly to that for when the variance is known only to lie in some finite range. The result of including the entropic measure factor is to suggest a spectrum consistent with the variance of the data which has less structure than that given by the forward transform. An application of the methodology to example data is demonstrated.

💡 Research Summary

The paper presents a Bayesian maximum‑entropy (MaxEnt) framework for estimating the power spectrum of irregularly sampled time‑series data that may contain gaps. Traditional Fourier‑based spectral analysis assumes uniformly spaced samples; when this assumption is violated, the forward transform often produces spurious peaks and over‑interprets noise as genuine structure. To address this, the authors formulate a joint objective consisting of a Gaussian likelihood term (derived from the data covariance matrix) and an entropy regularizer that penalises overly complex spectra.

A central contribution is the demonstration that the customary arbitrary prefactor multiplying the entropy term is unnecessary. By adopting a continuous Poisson prior for the amplitude coefficients of the Fourier basis, the entropy naturally takes the form (-\sum_i |x_i|) (where (x_i) are complex amplitudes), eliminating the need for an ad‑hoc scaling parameter. This choice ensures that the regularizer is properly normalised and that the resulting spectrum is driven solely by the data and the intrinsic prior information.

The authors treat two distinct scenarios: (1) the noise variance (or full covariance matrix) is known precisely, allowing a straightforward likelihood; (2) the variance is only known to lie within a finite interval, in which case it is promoted to a hyper‑parameter and integrated out using evidence maximisation. In both cases the optimisation proceeds via a Newton‑Raphson scheme applied to the combined likelihood‑entropy functional, with the irregular sampling matrix handled through least‑squares techniques rather than explicit matrix inversion. Missing data points are accommodated by inflating the corresponding entries of the covariance matrix, effectively down‑weighting those observations.

Synthetic experiments illustrate that, compared with the raw Fourier transform, the MaxEnt spectrum exhibits markedly reduced high‑frequency noise and fewer artificial peaks, while still recovering the true underlying frequencies. The method is then applied to an actual stellar luminosity record containing irregular observation times and gaps. The MaxEnt spectrum clearly isolates the known pulsation frequency of the star, suppresses unrelated high‑frequency components, and yields a smooth background consistent with the measured variance. Even when the exact variance is uncertain, the approach produces a spectrum that remains stable within the prescribed variance bounds.

In summary, the paper shows that a properly constructed entropy term—derived from a continuous Poisson prior—provides an effective regularisation that prevents over‑interpretation of imperfect data. The MaxEnt framework delivers spectra that are both statistically consistent with the data’s noise characteristics and physically interpretable, making it a valuable tool for astronomical time‑series analysis and any other domain where irregular sampling and data gaps are common.