Anti-crosstalk

Getting started

As the concepts here are quite new to us, the first thing we should do is cook up some super simple synthetic examples. With working synthetic codes in hand we should see if we can go ahead and repair the Galilee survey. Hopefully we'll recognize we have built some reusable software to facilitate other projects.

Getting started will not be easy. Most commonly we have a simple synthetic example under control and struggle to find an appropriate real data set. Here we have a suitable beginners' data set (Galilee) but we need to find a synthetic data set to provide examples that give clarity to the whole process. Just one issue is dealing with the relative scaling of the three regularizations. We'd like meaningful examples where only one or two of the regularizations are actually required.

There are many paths to explore with anti-crosstalk technology. Besides the many potential applications one can hope that the anti-crosstalk regularization eliminates (or reduces) the need for the usual regularizations. That would be nice if true. The need to specify a suitable regularization is often what makes it difficult to automate data analysis based on inversion.

I'm worried about the job as I defined it for the first-year SEP students. Given the linearization I suggested to them, did they know how to measure success?

Anti-crosstalk