This *is* interesting.

And it also worries me, because there are only two possible values so there
really isn't that much spatial information to be had - however there is a
ton of sequential information. The problem is or I wonder if the temporal
detection can detect that after 1, 0, 0, 1, 0, 1, 1, 1, 0, 1 comes a "1" or
a "0" ? Where you see that "1" only follows a zero after 8 previous inputs
in the sequence have been inputted? Can this be done?

David

On Wed, Jan 28, 2015 at 7:27 PM, Chandan Maruthi <[email protected]>
wrote:

> Interesting.
> Theoritically.
> Either approach should work. Given enough data the system would find that
> there are only 2 unique low level patterns. And look for unique higher
> level patterns as a combination of low level patterns.
>
> Chandan
>
>
> On Thursday, January 29, 2015, Matthew Taylor <[email protected]> wrote:
>
>> I have a project where I'd like to send a stream of boolean values
>> into NuPIC, but it looks like boolean values are not a valid data type
>> for model input [1]. Should I use integers (0,1) to represent
>> booleans? Or should I use string categories (True,False)?
>>
>> [1] https://github.com/numenta/nupic/issues/1775
>>
>> Thanks,
>> ---------
>> Matt Taylor
>> OS Community Flag-Bearer
>> Numenta
>>
>>


-- 
*We find it hard to hear what another is saying because of how loudly "who
one is", speaks...*

Reply via email to