it's from python temporal pooler TP.py

On Wed, May 4, 2016 at 3:06 PM, Wakan Tanka <[email protected]> wrote:

> Weiru Zeng,
> May I ask from which source is the mentioned code? Thanks
>
> On Wed, May 4, 2016 at 1:27 PM, Marek Otahal <[email protected]> wrote:
>
>> 1/ CLA model is "the network", your description.py sets params for the
>> model;
>> 2/ OPF is framework, eg now for constructiong a model from the params and
>> running swarming; network describes how regions are connected (CLAmodel is
>> an example of a network)
>> 3/ no, it says the other direction of implication: if maxSegments.. are
>> not -1, global devay must be disabled.
>> 4/ you'll find and issue for that: fixed 'mode': number of synapses per
>> segment, segs per cell is given and fixed; "variable sized/growing" - these
>> can change as needed. AFAIK not supported now.
>>
>> Cheers,
>>
>>
>> On Wed, May 4, 2016 at 9:27 AM, Weiru Zeng <[email protected]> wrote:
>>
>>> Hi everyone:
>>>
>>> Recently, I learned the model_params.py of the opf. The code below which
>>> is in the TP.py confuses me:
>>>
>>>
>>> # Fixed size CLA mode?
>>> if maxSegmentsPerCell != -1 or maxSynapsesPerSegment != -1:
>>>   assert (maxSegmentsPerCell > 0 and maxSynapsesPerSegment > 0)
>>>   assert (globalDecay == 0.0)
>>>   assert (maxAge == 0)
>>>
>>>
>>> The question is:
>>>
>>> 1. What's the CLA model?
>>> 2. What's the relationship of the opf, clamodel and network?
>>> 3. Dose the code above means that maxSegmentsPerCell and
>>> maxSynapsesPerSegment must be set to -1 by default if golbalDecay is used?
>>>     On the contrary, if this two parameters great than 0, there will be
>>> no decay to the segments of cells that did not become active. Is it right?
>>> 4. What's the mean of the "fixed size CLA model"?
>>>
>>
>>
>>
>> --
>> Marek Otahal :o)
>>
>
>
>
> --
> Best Regards
>
> Name: Wakan Tanka a.k.a. Wakatana a.k.a. MackoP00h
> Location: Europe
> Note: I'm non native English speaker so please bare with me ;)
> Contact:
> [email protected]
> http://stackoverflow.com/users/1616488/wakan-tanka
> https://github.com/wakatana
> https://twitter.com/MackoP00h
>



-- 
Marek Otahal :o)

Reply via email to