Cones, hilbert metric, rates of contraction


http://www.math.uh.edu/~climenha/blog-posts/hilbert-metric-2.pdf

Refines markov chains to compute contraction rate.

Builds on ideas that right eigenket controls distribution evolution, left eigenket controls scaling, top eigenvalue of the operators spectrum induces convergence (towards connectedness, in the path model), and the second largest controls mixing rate.

Its funny how dirac notation, alone, doesnt make it intuitive that distributions themselves can act as a basis (coordinate system); but the cone model make it entirely intuitive.

One also gets to see however, relevant to nsa backdoors, how one can make the evolution react more strongly to certain input partitions. Even with sigaba, perhaps one should be prequalifying keys, for certain plaintext priors, choosing per classification only those keys that minimise the  effect. Or (if you are 1980s nsa wanting a backdoor) you leverage that bias. If you are 2013 nsa, you exploit the imposed-untrustworthiness of the American internet telco to induce the biases in end-end cryptosystems that you want, per type of channel.

This all feels just like tunny!

Advertisements

About home_pw@msn.com

Computer Programmer who often does network administration with focus on security servers. Very strong in Microsoft Azure cloud!
This entry was posted in coding theory. Bookmark the permalink.