Convergence of Levenberg--Marquardt method in Banach spaces

We study the Levenberg--Marquardt method for nonlinear inverse problems in Banach spaces under a Hölder stability estimate. Under suitable assumptions on the geometry of the Banach space and the Fréchet derivative of the measurement operator, we establish the well-definedness of the method, local convergence and convergence rates for exact data, and the monotonic decrease of the Bregman distance up to the stopping index for noisy data. These results extend the corresponding Hilbert space theory to the Banach space setting.