Re: [Pcsim-users] How to make a field recordable?

SourceForge Headquarters 225 Broadway Suite 1600 San Diego, CA 92101 +1 (858) 422-6466

[Sorry I forgot to send the response to the list]

Hi Dejan,

Thanks to your remark I have realised my error! In a previous version I 
used "linear_sum" and wrote the python with this name. I later changed 
the C++ member to "linear_input" but not the python code. It was a 
stupid mistake :(

I have updated the code and I can now record the field.

However, I would like to create a member function double 
getLinearInput(void) that returns the value of the member "linear_input" 
(in the same vein as getAnalogOutput()). How do I have to declare it?

I have tried the following declaration double getLinearInput(void) { 
return this->linear_input;}; (see attachments) but when I try to use it 
I get the following error:

ArgumentError: Python argument types in

    SquashingNeuron.getLinearInput(SquashingNeuron)

did not match C++ signature:

    getLinearInput(SquashingNeuron {lvalue})

I have seen that AnalogPointNeuron::getAnalogOutput is declared as follows:

virtual double <http://www.lsm.tugraz.at/pcsim/cppclassreference/html/classdouble.html> getAnalogOutput <http://www.lsm.tugraz.at/pcsim/cppclassreference/html/classAnalogPointNeuron.html#a6e02306ef9cd057f844615691b9293f> (analog_port_id_t <http://www.lsm.tugraz.at/pcsim/cppclassreference/html/globaldefinitions_8h.html#40dfc683bdf73c81c9c73726fc834feb> port=0) const

Is it necessary to declare the function virtual (I'm not a C++ expert :) ?

Thank you again & sorry for the trouble,
Mathieu

Pecevski Dejan wrote:
> Hi Mathieu,
>
> I've looked at your code.
> Do you mean the linear_input public member in the SquashingNeuron 
> function, instead of linear_sum?
> Did you try the linear_input member or linear_sum?
>
> regards,
> Dejan
>
>
> Mathieu Dubois wrote:
>> Hi everybody,
>>
>> I'm developing a small PCSIM extension (I have carefully followed the 
>> steps in the HowTo).
>>
>> I want to define a new class SquashingNeuron which derives from 
>> LinearNeuron. At each step it should sum the synaptic inputs (like 
>> LinearNeuron) and apply a squashing function.
>>
>> The implementation is straightforward (see attachments). Compilation 
>> & tests went fine.
>>
>> However (for debugging purpose) I would like to be able to record the 
>> "linear_sum" public member (which is the linear sum of synaptic 
>> inputs - result of LinearNeuron::advance(...) see code).
>>
>> When I try this I get the following error message:
>>
>> RuntimeError: PCSIM::Exception in function SimObject::findField: 
>> SquashingNeuron has no field named linear_sum
>>
>> Is there something wrong with my code or do I have to do something to 
>> turn a C++ public member into a PCSIM field?
>>
>> Interested readers will recognize the kind of squashing function used 
>> in the p-delta learning rule... which is what I would like to implement.
>>
>> Thanks in advance,
>> Mathieu
>> ------------------------------------------------------------------------
>>
>> #include "SquashingNeuron.h"
>>
>> // SquashingNeuron::SquashingNeuron(): rho(DEFAULT_RHO) {
>> // }
>>
>> SquashingNeuron::SquashingNeuron(double rho): rho(rho) {
>> }
>>
>> int SquashingNeuron::advance(AdvanceInfo const & info) {
>>     LinearNeuron::advance(info);
>>     linear_input = this->Vm;
>>     if (linear_input < -this->rho) {
>>         this->Vm = POSITIVE_OUTPUT;
>>     }
>>     if ((linear_input >= -this->rho) && (linear_input <= this->rho)) {
>>          this->Vm = linear_input/this->rho;
>>     }
>>     if (linear_input > this->rho) {
>>         this->Vm = NEGATIVE_OUTPUT;
>>     }
>>     return 0;
>> }
>>   
>> ------------------------------------------------------------------------
>>
>> #ifndef SQUASHINGNEURON_H_
>> #define SQUASHINGNEURON_H_
>>
>> #include "LinearNeuron.h"
>>
>> #define POSITIVE_OUTPUT (1)
>> #define NEGATIVE_OUTPUT (-1)
>>
>> #define DEFAULT_RHO (0.25)
>>
>> class SquashingNeuron : public LinearNeuron
>> {
>>
>>     SIMOBJECT( SquashingNeuron, AdvancePhase::Two )
>>
>> public:
>> //     SquashingNeuron();
>>     SquashingNeuron(double rho = DEFAULT_RHO);
>>
>>     int advance(AdvanceInfo const &);
>>
>>     double linear_input;
>>
>> protected:
>>     double rho;
>> };
>>
>> #endif /*SQUASHINGNEURON_H_*/
>>   
>> ------------------------------------------------------------------------
>>
>> ------------------------------------------------------------------------------
>> Apps built with the Adobe(R) Flex(R) framework and Flex Builder(TM) are
>> powering Web 2.0 with engaging, cross-platform capabilities. Quickly and
>> easily build your RIAs with Flex Builder, the Eclipse(TM)based development
>> software that enables intelligent coding and step-through debugging.
>> Download the free 60 day trial. http://p.sf.net/sfu/www-adobe-com
>> ------------------------------------------------------------------------
>>
>> _______________________________________________
>> Pcsim-users mailing list
>> Pcs...@li...
>> https://lists.sourceforge.net/lists/listinfo/pcsim-users
>>   
>
>
> -- 
> _______________________________________________
> Dejan Pecevski
> Institute for Theoretical Computer Science
> Graz University of Technology
> Inffeldgasse 16b, A-8010 Graz, Austria
> Tel. +43 316 873 5849
>   
> ------------------------------------------------------------------------
>
> ------------------------------------------------------------------------------
> Apps built with the Adobe(R) Flex(R) framework and Flex Builder(TM) are
> powering Web 2.0 with engaging, cross-platform capabilities. Quickly and
> easily build your RIAs with Flex Builder, the Eclipse(TM)based development
> software that enables intelligent coding and step-through debugging.
> Download the free 60 day trial. http://p.sf.net/sfu/www-adobe-com
> ------------------------------------------------------------------------
>
> _______________________________________________
> Pcsim-users mailing list
> Pcs...@li...
> https://lists.sourceforge.net/lists/listinfo/pcsim-users
>   