Daniela Calvetti

Daniela Calvetti, Salvatore Cuomo, Monica Pragliola et al.

The Linear Multistep Method Particle Filter (LMM PF) is a method for predicting the evolution in time of a evolutionary system governed by a system of differential equations. If some of the parameters of the governing equations are unknowns, it is possible to organize the calculations so as to estimate them while following the evolution of the system in time. The underlying assumption in the approach that we present is that all unknowns are modelled as random variables, where the randomness is an indication of the uncertainty of their values rather than an intrinsic property of the quantities. Consequently, the states of the system and the parameters are described in probabilistic terms by their density, often in the form of representative samples. This approach is particularly attractive in the context of parameter estimation inverse problems, because the statistical formulation naturally provides a means of assessing the uncertainty in the solution via the spread of the distribution. The computational efficiency of the underlying sampling technique is crucial for the success of the method, because the accuracy of the solution depends on the ability to produce representative samples from the distribution of the unknown parameters. In this paper LMM PF is tested on a skeletal muscle metabolism problem, which was previously treated within the Ensemble Kalman filtering framework. Here numerical evidences are used to highlight the correlation between the main sources of errors and the influence of the linera multistep method adopted. Finally, we analyzed the effect of replacing LMM with Runge-Kutta class integration methods for supporting the PF technique.

Daniela Calvetti, Erkki Somersalo

A way to lower computational cost in large scale inverse problems and problems depending on poorly known model parameters is to replace the detailed model by an approximate one. Inverse problems are typically ill-posed, and the model discrepancy introduced by using approximate models often shows up in the computed solutions as disturbing artifacts or blurring. In this article, we consider two methods of addressing certain types of modeling errors, the Bayesian approximation error (BAE) method and linear algebraic spotlight inversion to suppress clutter in the computational model by orthogonal projections. Through the process of analyzing the two approaches, we show that they turn out to be closely related but not equivalent, and we highlight a connection to sketching schemes in randomized linear algebra. The similarities between the methods and their successful suppression of most of the clutter effects is elucidated with two computed examples, one addressing of X-ray tomography and the other electrical impedance tomography.

Alberto Bocchinfuso, Daniela Calvetti, Erkki Somersalo

Dictionary learning methods continue to gain popularity for the solution of challenging inverse problems. In the dictionary learning approach, the computational forward model is replaced by a large dictionary of possible outcomes, and the problem is to identify the dictionary entries that best match the data, akin to traditional query matching in search engines. Sparse coding techniques are used to guarantee that the dictionary matching identifies only few of the dictionary entries, and dictionary compression methods are used to reduce the complexity of the matching problem. In this article, we propose a work flow to facilitate the dictionary matching process. First, the full dictionary is divided into subdictionaries that are separately compressed. The error introduced by the dictionary compression is handled in the Bayesian framework as a modeling error. Furthermore, we propose a new Bayesian data-driven group sparsity coding method to help identify subdictionaries that are not relevant for the dictionary matching. After discarding irrelevant subdictionaries, the dictionary matching is addressed as a deflated problem using sparse coding. The compression and deflation steps can lead to substantial decreases of the computational complexity. The effectiveness of compensating for the dictionary compression error and using the novel group sparsity promotion to deflate the original dictionary are illustrated by applying the methodology to real world problems, the glitch detection in the LIGO experiment and hyperspectral remote sensing.

Daniela Calvetti, Annalisa Pascarella, Francesca Pitolli et al.

A recently proposed IAS MEG inverse solver algorithm, based on the coupling of a hierarchical Bayesian model with computationally efficient Krylov subspace linear solver, has been shown to perform well for both superficial and deep brain sources. However, a systematic study of its sensitivity and specificity as a function of the activity location is still missing. We propose novel statistical protocols to quantify the performance of MEG inverse solvers, focusing in particular on their sensitivity and specificity in identifying active brain regions. We use these protocols for a systematic study of the sensitivity and specificity of the IAS MEG inverse solver, comparing the performance with three standard inversion methods, wMNE, dSPM, and sLORETA. To avoid the bias of anecdotal tests towards a particular algorithm, the proposed protocols are Monte Carlo sampling based, generating an ensemble of activity patches in each brain region identified in a given atlas. The sensitivity is measured by how much, on average, the reconstructed activity is concentrated in the brain region of the simulated active patch. The specificity analysis is based on Bayes factors, interpreting the estimated current activity as data for testing the hypothesis that the active brain region is correctly identified, vs. the hypothesis of any erroneous attribution. The methodology allows the presence of a single or several simultaneous activity regions, without assuming the knowledge of the number of active regions. The testing protocols suggest that the IAS solver performs well in terms of sensitivity and specificity both with cortical and subcortical activity estimation.

Daniela Calvetti, Francesca Pitolli, Erkki Somersalo et al.

The solution of linear inverse problems when the unknown parameters outnumber data requires addressing the problem of a nontrivial null space. After restating the problem within the Bayesian framework, a priori information about the unknown can be utilized for determining the null space contribution to the solution. More specifically, if the solution of the associated linear system is computed by the Conjugate Gradient for Least Squares (CGLS) method, the additional information can be encoded in the form of a right preconditioner. In this paper we study how the right preconditioned changes the Krylov subspaces where the CGLS iterates live, and draw a tighter connection between Bayesian inference and Krylov subspace methods. The advantages of a Krylov-meet-Bayes approach to the solution of underdetermined linear inverse problems is illustrated with computed examples.