New AI method recovers corrupted signals with far fewer samples than current methods
Researchers have developed a Bayesian approach that uses AI-generated priors to reconstruct scrambled signals from significantly fewer noisy observations than existing methods require. The breakthrough could accelerate medical imaging analysis, particularly in cryo-electron microscopy, by reducing data collection costs and processing time—a major efficiency gain for labs and biotech companies relying on these techniques.
Originaltitel: Moment-Based Posterior Sampling for Multi-Reference Alignment
We propose a Bayesian approach to the problem of multi-reference alignment -- the recovery of signals from noisy, randomly shifted observations. While existing frequentist methods accurately recover the signal at arbitrarily low signal-to-noise ratios, they require a large number of samples to do so. In contrast, our proposed method leverages diffusion models as data-driven plug-and-play priors, conditioning these on the sample power spectrum (a shift-invariant statistic) enabling both accurate posterior sampling and uncertainty quantification. The use of an appropriate prior significantly reduces the required number of samples, as illustrated in simulation experiments with comparisons to state-of-the-art methods such as expectation--maximization and bispectrum inversion. These findings establish our approach as a promising framework for other orbit recovery problems, such as cryogenic electron microscopy (cryo-EM).