RE: Higher order representation and associative memory
> (Aseem Hegshetye)
Yeh you've got the right idea, you could actually go record two people speaking
and then youd be able to get responses either way.Which is pretty amazing,
thats one thing id definitely like to implement. (but note I don't use nupic,
but i'm interested in it, I cross htmwith more traditional ann's, in my
thinking.)
I don't know how good nupic is, but I imagine its pretty good, since its got
the main man behind it, Jeff Hawkins, and I trust his thinking.
I think of htm, or nets like it, as the ultimate compressor recorder, they have
a huge store, because they reuse parts of distributions thatcome from more than
one whole distribution. (a whole distribution being a whole picture of all the
inputs at one instant of time.)
How good it actually is, when one is developing unsupervised motor, I don't
know, but it may be pretty good, i'm not sure, sometimes Iget in silly moods
and then it seems like it actually really works well! :)
Its just unproven at this point, and all I can do is speculate.
_______________________________________________
nupic mailing list
[email protected]
http://lists.numenta.org/mailman/listinfo/nupic_lists.numenta.org
