top of page

Tikhonov regularization adds an incommensurable advantage to the ungrateful task of represent the interactions of the greatest number of variables in the hope to get a slyest representation of reality. P. S. Bear in mind that we target models “that are no more complicated than they need to be”.

In one hand, it saves SVD of its inability to represent [expert] knowledge constrains in the simple form of a unique coveriance matrix C(k)= σ2kI.  

A more 
reliable weighting strategy
...
SVD and Tikhonov

65.PNG
Tikhonov regularization 2

In other hand, Tikhonov regularization strategy ( 0 | 1 ) proposes to face, straightforwardly, the errors and noises as model to measurement misfit (Qm), and inherent imperfections of the model, simplifications and calibration constrains (Qr).

62.PNG
64.PNG

Beta μ

Coming soon: A detailed and practical review of how to construct a weighting matrix (Q).
bottom of page