{"id":"arxiv/2603.14547","url":"https://arxiv.org/abs/2603.14547","tags":["math.NA","cs.IT","math.ST"],"score":0,"title":"Maximum Entropy Least Squares Solutions of Overdetermined Linear Systems","domain":"math.NA","source":"arxiv","authors":["Felice Iavernaro","Monica Lazzo","Lorenzo Pisani"],"reviews":0,"subject":"math.NA","summary":"We investigate the theoretical foundations of a recently introduced entropy-based formulation of weighted least squares for the approximation of overdetermined linear systems, motivated by robust data fitting in the presence of sparse gross errors. The weight vector is interpreted as a discrete probability distribution and is determined by maximizing Shannon entropy under normalization and a prescribed mean squared error (MSE) constraint. Unlike classical ordinary least squares, where the error level is an output of the minimization process, here the MSE value plays the role of a control…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-03-15","reactions":{}}