Non-Parametric Bayesian Framework Resolves Computational Bias Debates
CSIRO and the Event Horizon Telescope (EHT) collaboration published in Nature on November 11, 2026, a re-analysis of historical millimeter interferometry observations of M87* using a model-independent image reconstruction algorithm; the black hole’s shadow and radiant accretion ring were re-confirmed. After the 2019 first black hole image, independent researchers raised questions about image deconvolution and algorithmic bias. The CSIRO-led consortium developed a non-parametric Bayesian imaging algorithm without predefined morphological priors, performing blind statistical reconstruction from raw Very Long Baseline Interferometry (VLBI) visibilities.
According to Nature and ScienceDaily, CSIRO Lead Research Scientist Dr. David Rayner discussed the breakthrough during a media symposium. Rayner emphasized: “Our algorithm removes subjective geometric constraints from the reconstruction process. Working strictly from first-principles statistical physics, the reconstructed visibilities unequivocally demonstrate the bright photon ring enclosing the central black hole silhouette. This provides definitive confirmation of the original astrophysical discovery.”
Mathematical fitting established the photon ring diameter spans 42 microarcseconds, with azimuthal brightness asymmetry matching relativistic Doppler beaming predictions from Einstein’s general theory of relativity. The research synthesized recalibrated data from ALMA in Chile and SPT in Antarctica; integrating high-gain southern stations improved interferometric signal-to-noise ratios by more than 35 percent.
Under the next-generation EHT (ngEHT) upgrade program, observational frequencies will expand from 230 GHz (1.3 mm) to 345 GHz (0.87 mm), sharpening angular resolution by an additional factor of two. High-precision geometric mapping of the M87* event horizon allows physicists to constrain black hole spin parameters and test deviations from the Kerr metric. Research teams are applying the validated Bayesian imaging software to re-analyze observational archives of Sagittarius A* (Sgr A*).
