"""
Current source classes for the neuron module.
Classes:
DCSource -- a single pulse of current of constant amplitude.
StepCurrentSource -- a step-wise time-varying current.
NoisyCurrentSource -- a Gaussian whitish noise current.
ACSource -- a sine modulated current.
:copyright: Copyright 2006-2020 by the PyNN team, see AUTHORS.
:license: CeCILL, see LICENSE for details.
"""
from neuron import h
import numpy
from pyNN.standardmodels import electrodes, build_translations, StandardCurrentSource
from pyNN.parameters import ParameterSpace, Sequence
from pyNN.neuron import simulator
class NeuronCurrentSource(StandardCurrentSource):
"""Base class for a source of current to be injected into a neuron."""
def __init__(self, **parameters):
self._devices = []
self.cell_list = []
self._amplitudes = None
self._times = None
self._h_iclamps = {}
parameter_space = ParameterSpace(self.default_parameters,
self.get_schema(),
shape=(1,))
parameter_space.update(**parameters)
parameter_space = self.translate(parameter_space)
self.set_native_parameters(parameter_space)
simulator.state.current_sources.append(self)
@property
def _h_amplitudes(self):
if self._amplitudes is None:
if isinstance(self.amplitudes, Sequence):
self._amplitudes = h.Vector(self.amplitudes.value)
else:
self._amplitudes = h.Vector(self.amplitudes)
return self._amplitudes
@property
def _h_times(self):
if self._times is None:
if isinstance(self.times, Sequence):
self._times = h.Vector(self.times.value)
else:
self._times = h.Vector(self.times)
return self._times
def _reset(self):
if self._is_computed:
self._amplitudes = None
self._times = None
self._generate()
for iclamp in self._h_iclamps.values():
self._update_iclamp(iclamp, 0.0) # send tstop = 0.0 on _reset()
def _update_iclamp(self, iclamp, tstop):
if not self._is_playable:
iclamp.delay = self.start
iclamp.dur = self.stop - self.start
iclamp.amp = self.amplitude
if self._is_playable:
iclamp.delay = 0.0
iclamp.dur = 1e12
iclamp.amp = 0.0
# check exists only for StepCurrentSource (_is_playable = True, _is_computed = False)
# t_stop should be part of the time sequence to handle repeated runs
if not self._is_computed and tstop not in self._h_times.to_python():
ind = self._h_times.indwhere(">=", tstop)
if ind == -1: # tstop beyond last specified time instant
ind = self._h_times.size()
if ind == 0.0: # tstop before first specified time instant
amp_val = 0.0
else:
amp_val = self._h_amplitudes.x[int(ind)-1]
self._h_times.insrt(ind, tstop)
self._h_amplitudes.insrt(ind, amp_val)
self._h_amplitudes.play(iclamp._ref_amp, self._h_times)
def _check_step_times(self, times, amplitudes, resolution):
# ensure that all time stamps are non-negative
if not (times >= 0.0).all():
raise ValueError("Step current cannot accept negative timestamps.")
# ensure that times provided are of strictly increasing magnitudes
dt_times = numpy.diff(times)
if not all(dt_times>0.0):
raise ValueError("Step current timestamps should be monotonically increasing.")
# map timestamps to actual simulation time instants based on specified dt
for ind in range(len(times)):
times[ind] = self._round_timestamp(times[ind], resolution)
# remove duplicate timestamps, and corresponding amplitudes, after mapping
step_times = []
step_amplitudes = []
for ts0, amp0, ts1 in zip(times, amplitudes, times[1:]):
if ts0 != ts1:
step_times.append(ts0)
step_amplitudes.append(amp0)
step_times.append(times[-1])
step_amplitudes.append(amplitudes[-1])
return step_times, step_amplitudes
def set_native_parameters(self, parameters):
parameters.evaluate(simplify=True)
for name, value in parameters.items():
if name == "amplitudes": # key used only by StepCurrentSource
step_times = parameters["times"].value
step_amplitudes = parameters["amplitudes"].value
step_times, step_amplitudes = self._check_step_times(step_times, step_amplitudes, simulator.state.dt)
parameters["times"].value = step_times
parameters["amplitudes"].value = step_amplitudes
if isinstance(value, Sequence): # this shouldn't be necessary, but seems to prevent a segfault
value = value.value
object.__setattr__(self, name, value)
self._reset()
def get_native_parameters(self):
return ParameterSpace(dict((k, self.__getattribute__(k)) for k in self.get_native_names()))
def inject_into(self, cells):
__doc__ = StandardCurrentSource.inject_into.__doc__
for id in cells:
if id.local:
if not id.celltype.injectable:
raise TypeError("Can't inject current into a spike source.")
if not (id in self._h_iclamps):
self.cell_list += [id]
self._h_iclamps[id] = h.IClamp(0.5, sec=id._cell.source_section)
self._devices.append(self._h_iclamps[id])
def record(self):
self.itrace = h.Vector()
self.itrace.record(self._devices[0]._ref_i)
self.record_times = h.Vector()
self.record_times.record(h._ref_t)
def _get_data(self):
# NEURON and pyNN have different concepts of current initiation times
# To keep this consistent across simulators, pyNN will have current
# initiating at the electrode at t_start and effect on cell at next dt.
# This requires removing the first element from the current Vector
# as NEURON computes the currents one time step later. The vector length
# is compensated by repeating the last recorded value of current.
t_arr = numpy.array(self.record_times)
i_arr = numpy.array(self.itrace)[1:]
i_arr = numpy.append(i_arr, i_arr[-1])
return (t_arr, i_arr)
[docs]class DCSource(NeuronCurrentSource, electrodes.DCSource):
__doc__ = electrodes.DCSource.__doc__
translations = build_translations(
('amplitude', 'amplitude'),
('start', 'start'),
('stop', 'stop')
)
_is_playable = False
_is_computed = False
[docs]class StepCurrentSource(NeuronCurrentSource, electrodes.StepCurrentSource):
__doc__ = electrodes.StepCurrentSource.__doc__
translations = build_translations(
('amplitudes', 'amplitudes'),
('times', 'times')
)
_is_playable = True
_is_computed = False
def _generate(self):
pass
[docs]class ACSource(NeuronCurrentSource, electrodes.ACSource):
__doc__ = electrodes.ACSource.__doc__
translations = build_translations(
('amplitude', 'amplitude'),
('start', 'start'),
('stop', 'stop'),
('frequency', 'frequency'),
('offset', 'offset'),
('phase', 'phase')
)
_is_playable = True
_is_computed = True
def __init__(self, **parameters):
NeuronCurrentSource.__init__(self, **parameters)
self._generate()
def _generate(self):
# Not efficient at all... Is there a way to have those vectors computed on the fly ?
# Otherwise should have a buffer mechanism
temp_num_t = int(round(((self.stop + simulator.state.dt) - self.start) / simulator.state.dt))
tmp = simulator.state.dt * numpy.arange(temp_num_t)
self.times = tmp + self.start
self.amplitudes = self.offset + self.amplitude * numpy.sin(tmp * 2 * numpy.pi * self.frequency / 1000. + 2 * numpy.pi * self.phase / 360)
self.amplitudes[-1] = 0.0
[docs]class NoisyCurrentSource(NeuronCurrentSource, electrodes.NoisyCurrentSource):
__doc__ = electrodes.NoisyCurrentSource.__doc__
translations = build_translations(
('mean', 'mean'),
('start', 'start'),
('stop', 'stop'),
('stdev', 'stdev'),
('dt', 'dt')
)
_is_playable = True
_is_computed = True
def __init__(self, **parameters):
NeuronCurrentSource.__init__(self, **parameters)
self._generate()
def _generate(self):
## Not efficient at all... Is there a way to have those vectors computed on the fly ?
## Otherwise should have a buffer mechanism
temp_num_t = int(round((self.stop - self.start) / max(self.dt, simulator.state.dt)))
self.times = self.start + max(self.dt, simulator.state.dt) * numpy.arange(temp_num_t)
self.times = numpy.append(self.times, self.stop)
self.amplitudes = self.mean + self.stdev * numpy.random.randn(len(self.times))
self.amplitudes[-1] = 0.0
