[Bug] Synchrotool failes to annotate spikes if spike time == spiketrain.t_stop

[jo460464]

Describe the bug

Spiketrains can contain spikes with spiketimes <= spiketrain.t_stop. The Synchrotool can only annotate Spikes with a spike time <= spiketrain.t_stop - (0.5/sampling_rate). If the spike time is larger you get an IndexError.

To Reproduce

1. Create a neo.spiketrain with a spike at t_stop
2. Access the Sychrotool
3. Use Synchrotool.annotate_synchrofacts() or Synchrotool.delete_synchrofacts()

from elephant.spike_train_synchrony import Synchrotool
import quantities as pq
import neo

sts = [neo.SpikeTrain([0., 0.5, 1.0] *pq.s, t_start=0*pq.s, t_stop=1.*pq.s),
       neo.SpikeTrain([0.1, 0.5, 0.9] *pq.s, t_start=0*pq.s, t_stop=1.*pq.s)]
synch= Synchrotool(sts, sampling_rate=30*pq.kHz)
synch.annotate_synchrofacts()

Expected behavior

Annotate all spikes with the correct complexity

Environment

• OS: Linux, Ubuntu 22.04.3
• How you installed elephant: conda
• Python version: 3.12.0
• neo python package version: 0.12.0
• elephant python package version: 0.13.0

[Moritz-Alexander-Kern]

Hi,
thanks for the report.
I believe this is very similar to Issue NeuralEnsemble#493 .

[skrausse]

Hi, thanks for the report. I believe this is very similar to Issue NeuralEnsemble#493 .

Indeed it is, @jo460464 already found the relevant code in the complexity class, where the t_stop bin is excluded and therefore leads to a failure.

    def _epoch_no_spread(self):
        """
        Get an epoch object of the complexity distribution with `spread` = 0
        """
        left_edges = self.time_histogram.times
        durations = self.bin_size * np.ones(self.time_histogram.shape)

        if self.sampling_rate:
            # ensure that spikes are not on the bin edges
            bin_shift = .5 / self.sampling_rate
            left_edges -= bin_shift

            # Ensure that an epoch does not start before the minimum t_start.
            # Note: all spike trains share the same t_start and t_stop.
            if left_edges[0] < self.t_start:
                left_edges[0] = self.t_start
                durations[0] -= bin_shift
        else:
            warnings.warn('No sampling rate specified. '
                          'Note that using the complexity epoch to get '
                          'precise spike times can lead to rounding errors.')

        complexity = self.time_histogram.magnitude.flatten()
        complexity = complexity.astype(np.uint16)

        epoch = neo.Epoch(left_edges,
                          durations=durations,
                          array_annotations={'complexity': complexity})
        return epoch

Shifting the bins by half a sampling period leaves events at t_stop to be out of bounds.

Jonas and I will start working on a fix which we could maybe discuss next elephant meeting