testsuite::test_poisson_ps_intervals - checks coefficient of variationSynopsis:
(test_poisson_ps_intervals) run -> CV sufficiently close to 1Description:
The inter spike interval (ISI) distribution of a Poisson process is
exponential. The coefficient of variation defined as the standard deviation
of the intervals normalized by the mean interval is 1. The test checks whether
the output of the poisson_generator_ps as recorded by the spike_detector is
sufficiently close to 1. Without the property /precise_times the error is larger
than 0.1 .
An improved version of this test should check whether the observed deviation from
unity is within the error bounds given by the number of spikes.
Even a correct implementation generates with low probability spike trains which
do not pass the test. The seed set in the default configuration of NEST avoids
this problem. Therefore, failure of this script indicates that the configuration
is not portable.
The test in r8067 worked even with ticket #157 unresolved. Only the presence of a
neuron (see test_spike_transmission_ps_iaf) exhibited the problem. Prior to the
fix of ticket #164 this test produced inconsistent results without specifying
/local_num_threads 1 for the kernel.