Hi Aseem,

I can see those now, thanks. Looks like I needed to be logged in as myself
to see anything.

I'll have a look at the code and report back.

My email was titled "Hackathon Task - Per-cell Dendrites" which was to do
two things:

1) Load the SP activations with the predictive potentials from the previous
step, thus promoting those columns.
2) Use feedforward dendrites per cell rather than per column for
fine-tuning learning the inputs.

It turns out this is much easier than it looks. I have pseudocode at the
bottom of that email which shows how it can be done with only a few lines
of extra code. If you'd like to test it, that'd be great.





On Mon, Nov 11, 2013 at 12:49 PM, Aseem Hegshetye <[email protected]>wrote:

> Hi Fergal,
>
> I did not set the access private. I tried the same link and I could see
> that code in two .py files.
> Please check again and let me know if you can see those files.
> I can also make a readme file, that will help understanding the code
> easier.
> Code is too small. just two files.
> I could not find ur email in friday's emailing list. I assume u are
> talking about having every cell being connected to the input bits instead
> of every column.
> That would be awesome. We should collaborate .
> I am also trying to see if there is something wrong with my connection
> matrix. I see that it repeats last two predictions.
>
> Regards
> Aseem Hegshetye
>
> _______________________________________________
> nupic mailing list
> [email protected]
> http://lists.numenta.org/mailman/listinfo/nupic_lists.numenta.org
>



-- 

Fergal Byrne

<http://www.examsupport.ie>Brenter IT
[email protected] +353 83 4214179
Formerly of Adnet [email protected] http://www.adnet.ie
_______________________________________________
nupic mailing list
[email protected]
http://lists.numenta.org/mailman/listinfo/nupic_lists.numenta.org

Reply via email to