> A surprise discovery: it seems that all of our recent tests have been > using the original code in .../trunk/demod64a.f90, with its exp(x) > symbol metrics. I tried switching back to your simple p1/psum, p2/psum > mnetrics computed in .../trunk/rsdtest/demod64b.f90, with much degraded > results. I did not try re-tuning things accordingly. >
I assume that you mean ../trunk/demod64b.f90 (not a). Really?! Shoot. I guess that the comparison between metrics needed to be done. Maybe I should do that next. The approach would be to switch metrics - run (with much degraded results) - and use the fort.40 results to bootstrap up to the “right” probabilities and see where we land. This could affect the mr2sym results as well. I’m still surprised - because I did my high resolution 2D histograms for both sets of metrics, and they looked very different. I was convinced that the 8x8 matrix in fort.40 looked more like the high-resolution histogram for the simple metrics. Huh. Steve k9an ------------------------------------------------------------------------------ _______________________________________________ wsjt-devel mailing list wsjt-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/wsjt-devel
