I believe the 'inferences' dictionary should have everything you need, at least
I've seen it having everything when I used MultistepInference mode.
Yes, you should backtrack after resetting, so you always train a new sequence
at the start of a sentence. Another way to think about it is, at every sentence
terminator, feed in the next n characters (to add more useful context) and then
a reset before continuing (this should simplify the logic).
On Sun, Nov 17, 2013 at 8:54 AM, Marek Otahal <[email protected]>
wrote:
> Hi Chetan,
> On Sun, Nov 17, 2013 at 5:42 PM, Chetan Surpur <[email protected]> wrote:
>> Yes, this is what I would suggest as well. I believe the OPF CLA model
>> already offers multiple possible inferences with a probability on each one.
>>
> I think it does. (btw, do you know how exactly to call it? or some doc?
> What we see now are values for multistep inference, but I need
> multiinference (one step) ) But I'm lazy, I want it to return the
> 'prediction' (which is field for "best" prediction) according to the
> probability distribution. If it doesn't do that, we should add it.
>>
>> Another thing you could try to make it work even better is, instead of
>> resetting the sequence as soon as you see a sentence terminating symbol,
>> set a timer to reset after seeing n more characters.
>>
> This is very good idea! "..so they sailed to the sea.<TERMINATOR> The next
> morn<RESET>" After the reset, should I backtrack to the terminator and
> re-feed the "The next morn.." again?
>> Then, your model will understand what comes after each type of sentence
>> terminator, and since it will include context about the characters near the
>> end of the sentence, the next sentence prediction won't be the same every
>> time.
>>
>>
> --
> Marek Otahal :o)
_______________________________________________
nupic mailing list
[email protected]
http://lists.numenta.org/mailman/listinfo/nupic_lists.numenta.org
