When a CLA Model returns a prediction object (in the OPF, like in the
hot gym tutorial), that prediction result object contains details
about exactly how confident it is about the prediction. It is actually
making several predictions at once, and each one is associated with a
confidence you can extract.

That is actually something I wanted to show in a future tutorial,
because it would be interesting to plot not only the "best
prediction", but the 2nd and 3rd best predictions and their associated
confidences, perhaps as heat signatures. Anyway, you can find out more
about the ModelResult object and how to get at this data on the wiki
[1]. This isn't documented extremely well, so you might want to just
dump the model result object so you can inspect its structure.

[1] 
https://github.com/numenta/nupic/wiki/Online-Prediction-Framework#model-output-the-modelresult-object

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


On Tue, Apr 29, 2014 at 3:14 PM, Mark Wilson <[email protected]> wrote:
> Hi
>
> Is there any sort of indicator to say how good NuPIC thinks a prediction is?
> The Anomaly score could be used, however I figure that it isn't really what
> it is for. Also the Anomoly score isn't available if you want to predict a
> number of steps ahead.
>
> Cheers
>
> Mark
>
>
> _______________________________________________
> nupic mailing list
> [email protected]
> http://lists.numenta.org/mailman/listinfo/nupic_lists.numenta.org
>

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

Reply via email to