Hey guys,

I was talking to "addos" on IRC, and a couple of questions came up
about swarming that I can't answer. Since I am also curious about
these, I'm asking them here!

#1. Does the errorScore reported during swarming give any indication
about how well a model will actually do?

For example, if one swarm on a data set returns a lower error score
than another swarm on a different data set, is that an indication that
the first model params should perform better than the 2nd? Or is the
error score relative to the performance of other models within the
swarm set?

#2. If I swarm for a one-step ahead prediction, but I want to switch
to a 5-step ahead prediction, should I reswarm?

Thanks,
---------
Matt Taylor
OS Community Flag-Bearer
Numenta

_______________________________________________
nupic mailing list
[email protected]
http://lists.numenta.org/mailman/listinfo/nupic_lists.numenta.org

Reply via email to