Computational Information Geometry Wonderland

@FrnkNlsn

Rayleigh mixture models (RMMs) are mixtures of exponential families (as Gaussian mixture models, GMMs). Here is a simple image where the foreground text has been synthesized with Rayleigh variates (lamda1) while the background is synthesized using Rayleigh variates (lamda2). I am curious to know how good human is at distinguishing distributions, compare to usual measures, say Kullback-Leibler divergence. Of course, it depends on the viewing distance/precision, otherwise we get expectation colors, and we have some work on this. Does stochasticity help in discriminating objects. I guess so.

Frank.

The jMEF library, Exponential families library, has been updated with new tutorials. A supplemental report listing formulas has been posted on arxiv: Statistical exponential families: A digest with flash cards

In statistics, exponential families generalize many well-known discrete and continuous distributions (Gaussian, Bernoulli, Poisson, Rayleigh), etc. Exponential families arose naturally from solving common sets of problems like the maximum entropy

More at http://en.wikipedia.org/wiki/Maximum_entropy.

That is again a short week-end post!