next up previous [pdf]

Next: EMPTY BINS AND INVERSE Up: Reproducible Documents

Regularization is model styling

Jon Claerbout

Regularization is a method used in mathematics and statistics to deal with insufficient information. The reader must supply additional information in the form of an operator. From where is this operator to come, and what does it mean? It amounts to us, as practitioners, specifying a ``style'' of model. Where the model is a signal or an image, it amounts to specifying one weighting function in physical space and another in Fourier space.