A problem about COBA

[spike bush]

Hello everyone ,

My network structure is as follows:

gIn->gExc,

there are 450 neurons in the Input group,and only one neuron in the gExc group .The synapse model I choose is COBA model . And the input frequency of input neurons is 10Hz ,which is to say the input neuron spikes at 100ms,200ms, 300ms,400ms,...and so on .However,the output neuron (gExc group) spikes at 120ms ,148ms and 181ms .And my question is ,In the first 199ms ,the input neurons only spike at 100ms,but the gExc spikes  three time between 100ms and 200ms .There is no input ,but the output neuron spikes . 

[kashyaph]

Hi, before trying to speculate on what is going on, can you please provide some more info:

1. Did you use the Poisson spike generator or the periodic spike generator(User guide 6.1.2)?

PeriodicSpikeGenerator PSG(rate, spikeAtZero);

If you used Poisson spike generator, if usually do not spike periodically.

2. Did you try to plot the raster plot of spike from each group? How do you know the exact spike times?

http://uci-carl.github.io/CARLsim4/ch9_matlab_oat.html

[kashyaph]

Hello, were you able to resolve the issue? If not, can you please upload your code for the network or a toy example showing the problem?

[kashyaph]

Hi Spike Bush,

You are instantiating new methods in your ConstantISI class that are not part of the original spikeGenerator class. I ran into this issue.

*****************************************************************************************************************************************************************
/usr/local/cuda/bin/nvcc -ccbin g++ -I/home/hirak/CARL/include -L/home/hirak/CARL/lib -Wno-deprecated-gpu-targets   src/main_spike_propagate.cpp -o spike_propagate -lcarlsim -lcurand
src/main_spike_propagate.cpp: In function ‘int main(int, const char**)’:
src/main_spike_propagate.cpp:154:14: error: cannot declare variable ‘constISI’ to be of abstract type ‘ConstantISI’
  ConstantISI constISI(imgDim.N);
              ^
src/main_spike_propagate.cpp:12:7: note:   because the following virtual functions are pure within ‘ConstantISI’:
 class ConstantISI : public SpikeGenerator {
       ^
In file included from /home/hirak/CARL/include/carlsim.h:62:0,
                 from src/main_spike_propagate.cpp:1:
/home/hirak/CARL/include/callback.h:79:14: note:     virtual int SpikeGenerator::nextSpikeTime(CARLsim*, int, int, int, int, int)
  virtual int nextSpikeTime(CARLsim* s, int grpId, int i, int currentTime, int lastScheduledSpikeTime, int endOfTimeSlice) = 0;
              ^
make: *** [spike_propagate] Error 1
*************************************************************************************************************************************************

To generate periodic spikes, you can simply do this.

#include <carlsim.h>

// configure network

CARLsim sim("PSG example",CPU_MODE,USER);

int g0 = sim.createSpikeGeneratorGroup("input", 10, EXCITATORY_NEURON);

// etc.

// create periodic spike gen

PeriodicSpikeGenerator PSG(50.0f, false); // second argument to whether spike/not spike at zero

// associate group g0 with PSG

sim.setSpikeGenerator(g0, &PSG);

sim.setupNetwork();

// etc.

So, using the same network as you (attached script) I did  not get any spike in the excitatory group. Try playing with the connect parameters as well as neuron parameters.

Also, the behavior you observed might be due to use of conductance based synapses. Conductance based synapses hold presynaptic action potential for a while and decays exponentilly. So a spike arriving at 100 ms might still hold high conductance values for a later time, say 180 ms.

You should read through our docs regarding conductance based synapses.

http://uci-carl.github.io/CARLsim4/ch3_neurons_synapses_groups.html#ch3s2s2_coba