def test_2_runs(self):
"""
Test setup for 2 cells and their records which should be the same.
After each creation of the Simulation object, the Record object is cleaned up
(inside vector is empty) however on other Hoc object is removed eg. Sections.
That means - each new sections and cell are retained in the current NEURON run.
"""
def run_and_get_rec():
rec = Record(self.soma(0.5))
iclamp = IClamp(segment=self.soma(0.5))
iclamp.stim(delay=25, dur=3, amp=3)
sim = Simulation(init_v=-70, warmup=20)
sim.run(100)
return rec
rec1 = run_and_get_rec()
r1 = rec1.as_numpy('v').records
self.tearDown()
self.setUp()
rec2 = run_and_get_rec()
r2 = rec2.as_numpy('v').records
sim = Simulation(init_v=-70, warmup=20)
sim.run(100)
# Make assertions
self.assertEqual(4011, r1.size)
self.assertEqual(4011, r2.size)
self.assertTrue(np.alltrue(r1 == r2))
def run_and_get_rec():
rec = Record(self.soma(0.5))
iclamp = IClamp(segment=self.soma(0.5))
iclamp.stim(delay=25, dur=3, amp=3)
sim = Simulation(init_v=-70, warmup=20)
sim.run(100)
return rec
def test_empty_method(self):
sim = Simulation()
self.assertFalse(sim.is_neuron_empty())
self.soma.remove_immediate_from_neuron()
self.apic1.remove_immediate_from_neuron()
self.cell.remove_immediate_from_neuron()
self.assertTrue(sim.is_neuron_empty())
def test_iclamp_after_sim(self):
"""
IClamp works in a regular way after simulation run. in this case (amplitude 3 pA)
we have spike at the soma.
However IClamp delay is in the absolute time.
"""
# Record
rec = Record(self.soma(0.5))
# Run
sim = Simulation(init_v=-70, warmup=20)
sim.run(1)
# IClamp to soma
iclamp = IClamp(segment=self.soma(0.5))
iclamp.stim(delay=25, dur=3, amp=3)
sim.run(100)
r = rec.as_numpy(variable="v")
sim.remove_immediate_from_neuron()
rec.remove_immediate_from_neuron()
iclamp.remove_immediate_from_neuron()
# Make assertions
self.assertEqual(4051, r.size)
self.assertEqual(34.3815, round(r.records.max(), 4))
self.assertEqual(-75.3247, round(r.records.min(), 4))
self.assertEqual(330, r.records.argmax())
# time in ms of max mV value
self.assertEqual(28, round(r.time[r.records.argmax()], 4))
def test_netstim_after_sim(self):
"""
NetStim created after simulation run has no effect, it won't go on this simulation at all.
However NetStim start is in the absolute time.
"""
# Record
rec = Record(self.soma(0.5))
# Run
sim = Simulation(init_v=-70, warmup=20)
sim.run(1)
# Netstim to synapse
stim = NetStimCell("stim").add_netstim(start=25, number=10, interval=2)
self.cell.add_synapse(source=stim, netcon_weight=0.5, mod_name="ExpSyn", delay=1,
seg=self.apic1(0.5))
sim.run(100)
r = rec.as_numpy(variable="v")
# Make assertions
self.assertEqual(4051, r.size)
self.assertEqual(-67.1917, round(r.records.max(), 4))
self.assertEqual(-70.0, round(r.records.min(), 4))
self.assertEqual(3, r.records.argmax())
# time in ms of max mV value
self.assertEqual(6, r.time[r.records.argmax()])
stim.remove_immediate_from_neuron()
sim.remove_immediate_from_neuron()
rec.remove_immediate_from_neuron()
def test_netcon_event_after_sim(self):
syn = self.cell.add_synapse(source=None, netcon_weight=1.0, mod_name="ExpSyn", delay=1,
seg=self.apic1(0.5))
# Record
rec = Record(self.soma(0.5))
# Run
sim = Simulation(init_v=-70, warmup=20)
sim.run(1)
syn.make_event(50)
sim.run(100)
r = rec.as_numpy(variable="v")
syn.remove_immediate_from_neuron()
sim.remove_immediate_from_neuron()
rec.remove_immediate_from_neuron()
# Make assertions
self.assertEqual(4051, r.size)
self.assertEqual(34.4582, round(r.records.max(), 4))
self.assertEqual(-75.3478, round(r.records.min(), 4))
self.assertEqual(2159, r.records.argmax())
# time in ms of max mV value
self.assertEqual(73.725, round(r.time[r.records.argmax()], 4))
def test_netstim_before_sim(self):
"""
NetStim run in a regular way before Simulation run, in this cale (netcon weight=1.0)
we have spike of the soma.
However NetStim start is in the absolute time.
"""
# Netstim to synapse
stim = NetStimCell("stim").add_netstim(start=25, number=10, interval=2)
self.cell.add_synapse(source=stim, netcon_weight=1.0, mod_name="ExpSyn", delay=1,
seg=self.apic1(0.5))
# Record
rec = Record(self.soma(0.5))
# Run
sim = Simulation(init_v=-70, warmup=20)
sim.run(1)
sim.run(100)
r = rec.as_numpy(variable="v")
stim.remove_immediate_from_neuron()
sim.remove_immediate_from_neuron()
rec.remove_immediate_from_neuron()
# Make assertions
self.assertEqual(4051, r.size)
self.assertEqual(34.5205, round(r.records.max(), 4))
self.assertEqual(-75.3053, round(r.records.min(), 4))
self.assertEqual(319, r.records.argmax())
# time in ms of max mV value
self.assertEqual(27.725, round(r.time[r.records.argmax()], 4))
def test_add_netcon_and_make_event_separately(self):
g1 = self.cell.filter_synaptic_group(mod_name="Exp2Syn", tag="aa")
nc = g1.add_netcon(source=None)
exp2syn_rec = Record(g1['Exp2Syn'][0], variables='i')
expsyn_rec = Record(g1['ExpSyn'][0], variables='i')
sim = Simulation()
sim.run(1)
nc['Exp2Syn'][0].make_event(10)
nc['ExpSyn'][0].make_event(20)
sim.run(100)
exp2syn_np = exp2syn_rec.as_numpy('i')
stim_time_exp2syn = exp2syn_np.time[(exp2syn_np.records != 0).argmax()]
expsyn_np = expsyn_rec.as_numpy('i')
stim_time_expsyn = expsyn_np.time[(expsyn_np.records != 0).argmax()]
self.assertEqual(12.05, round(stim_time_exp2syn, 4))
self.assertEqual(22.025, round(stim_time_expsyn, 4))
# Remove variables and clear NEURON
nc['Exp2Syn'][0].remove_immediate_from_neuron()
nc['ExpSyn'][0].remove_immediate_from_neuron()
nc = {}
g1.remove_immediate_from_neuron()
exp2syn_rec.remove_immediate_from_neuron()
expsyn_rec.remove_immediate_from_neuron()
sim.remove_immediate_from_neuron()
def test_netcon_event_before_sim(self):
"""
Netcon created before sim init or start is not effective and should return error
"""
sim = Simulation()
sim.reinit()
syn = self.cell.add_synapse(source=None, netcon_weight=1.0, mod_name="ExpSyn", delay=1,
seg=self.apic1(0.5))
error = False
try:
syn.make_event(50)
except ConnectionRefusedError:
error = True
self.assertTrue(error)
def test_dt(self):
rec = Record(self.soma(0.5))
sim = Simulation(dt=10)
sim.run(100)
r1 = rec.as_numpy('v')
sim = Simulation(dt=0.01)
sim.run(100)
r2 = rec.as_numpy('v')
sim.remove_immediate_from_neuron()
rec.remove_immediate_from_neuron()
self.assertEqual(11, r1.size)
self.assertEqual(10001, r2.size)
def test_init_v(self):
rec = Record(self.soma(0.5))
sim = Simulation(init_v=-100)
sim.run(100)
r1 = rec.as_numpy('v')
sim = Simulation(init_v=100)
sim.run(100)
r2 = rec.as_numpy('v')
sim.remove_immediate_from_neuron()
rec.remove_immediate_from_neuron()
self.assertEqual(-100, r1.records[0])
self.assertEqual(100, r2.records[0])
def test_size_method(self):
sim = Simulation()
self.assertEqual(198, sim.size)
self.soma.remove_immediate_from_neuron()
self.apic1.remove_immediate_from_neuron()
self.cell.remove_immediate_from_neuron()
self.assertEqual(0, sim.size)
def test_record_before_sim(self):
"""
Record created before simulation run is full of data
"""
# Record
rec = Record(self.soma(0.5))
# IClamp to soma
iclamp = IClamp(segment=self.soma(0.5))
iclamp.stim(delay=25, dur=3, amp=3)
# Run
sim = Simulation(init_v=-70, warmup=20)
sim.run(1)
print(h.t)
sim.run(100)
r = rec.as_numpy(variable="v")
# Make assertions
self.assertEqual(4051, r.size)
self.assertEqual(34.3815, round(r.records.max(), 4))
self.assertEqual(-75.3247, round(r.records.min(), 4))
self.assertEqual(330, r.records.argmax())
# time in ms of max mV value
self.assertEqual(28, round(r.time[r.records.argmax()], 4))
def test_stim_syns(self):
gs = self.cell.filter_synaptic_group()
gs0_exp2syn_rec = Record(gs[0]['Exp2Syn'][0], variables='i')
gs0_expsyn_rec = Record(gs[0]['ExpSyn'][0], variables='i')
gs1_exp2syn_rec = Record(gs[1]['Exp2Syn'][0], variables='i')
gs1_expsyn_rec = Record(gs[1]['ExpSyn'][0], variables='i')
sim = Simulation()
sim.run(1)
gs[0].make_event(10)
gs[1].make_event(20)
sim.run(100)
# Test stim time of synaptic group 1
gs0_exp2syn_np = gs0_exp2syn_rec.as_numpy('i')
stim_time_gs0_exp2syn = gs0_exp2syn_np.time[(gs0_exp2syn_np.records != 0).argmax()]
gs0_expsyn_np = gs0_expsyn_rec.as_numpy('i')
stim_time_gs0_expsyn = gs0_expsyn_np.time[(gs0_expsyn_np.records != 0).argmax()]
self.assertEqual(round(stim_time_gs0_exp2syn, 1), round(stim_time_gs0_expsyn, 1))
# Test stim time of synaptic group 2
gs1_exp2syn_np = gs1_exp2syn_rec.as_numpy('i')
stim_time_gs1_exp2syn = gs1_exp2syn_np.time[(gs1_exp2syn_np.records != 0).argmax()]
gs1_expsyn_np = gs1_expsyn_rec.as_numpy('i')
stim_time_gs1_expsyn = gs1_expsyn_np.time[(gs1_expsyn_np.records != 0).argmax()]
self.assertEqual(round(stim_time_gs1_exp2syn, 1), round(stim_time_gs1_expsyn, 1))
# Test values of mV in soma
self.assertEqual(31.3285, round(gs0_exp2syn_np.records.max(), 4))
self.assertEqual(-61.309, round(gs0_exp2syn_np.records.min(), 4))
# Remove variables and clear NEURON
gs0_exp2syn_rec.remove_immediate_from_neuron()
gs0_expsyn_rec.remove_immediate_from_neuron()
gs1_exp2syn_rec.remove_immediate_from_neuron()
gs1_expsyn_rec.remove_immediate_from_neuron()
gs[0].remove_immediate_from_neuron()
gs[1].remove_immediate_from_neuron()
sim.remove_immediate_from_neuron()
def init(self,
input_cells: List[Cell],
output_cells: List[Cell],
reward_syns: List[Synapse] = (),
punish_syns: List[Synapse] = (),
init_v=-70,
warmup=0,
dt=0.1):
"""
Before Agent step() you need to call:
1. agent.build()
2. agent.init()
:param input_cells:
list of input cells
:param output_cells:
list of output cells
:param init_v:
:param warmup:
:param dt:
"""
self.input_cells = input_cells
self.output_cells = output_cells
self.reward_syns = reward_syns
self.punish_syns = punish_syns
if self.sim is not None:
raise RuntimeError(
"Simulation cannot been run before initialization.")
self.warmup = warmup
self.sim = Simulation(init_v=init_v,
warmup=warmup,
dt=dt,
warmup_on_create=True)
def setUpClass(cls):
morpho_path = os.path.join(path, "..",
"commons/morphologies/asc/cell2.asc")
# Create cell
cell = Cell(name="cell")
cell.load_morpho(filepath=morpho_path)
cell.insert("pas")
cell.insert("hh")
soma = cell.filter_secs("soma")
dend = cell.filter_secs("apic[10]")
syn = cell.add_synapse(source=None, mod_name="ExpSyn", seg=dend(0.5))
# Prepare EPSP and AP (IClamp) protocols
experiment = Experiment(iti=40)
experiment.add_epsp(num=3,
synapse=syn,
init=20,
interval=20,
weight=0.02)
experiment.add_iclamp(num=3,
segment=soma(0.5),
init=60,
interval=20,
dur=3,
amp=1.6)
experiment.build()
# Prepare plots
rec = Record([soma(0.5), dend(0.5)], variables='v')
# Run
sim = Simulation(init_v=-70,
warmup=20,
with_neuron_gui=False,
constant_timestep=True)
sim.run(runtime=100)
cls.v_soma = rec.as_numpy('v', segment_name=soma(.5).name)
cls.v_apic = rec.as_numpy('v', segment_name=dend(.5).name)
syn.remove_immediate_from_neuron()
soma.remove_immediate_from_neuron()
dend.remove_immediate_from_neuron()
rec.remove_immediate_from_neuron()
experiment.remove_immediate_from_neuron()
cell.remove_immediate_from_neuron()
sim.remove_immediate_from_neuron()
cls.sections_left = len(list(h.allsec()))
def test_record_after_sim(self):
"""
record created after simulation run is empty
"""
# IClamp to soma
iclamp = IClamp(segment=self.soma(0.5))
iclamp.stim(delay=25, dur=3, amp=3)
# Run
sim = Simulation(init_v=-70, warmup=20)
sim.run(1)
print(h.t)
# Record
rec = Record(self.soma(0.5))
sim.run(100)
r = rec.as_numpy(variable="v")
# Make assertion
self.assertEqual(0, r.size)
ns_cell = NetStimCell("stim_cell")
ns = ns_cell.make_netstim(start=30, number=5, interval=10)
syns = cell.add_synapses_with_spine(source=ns,
secs=cell.filter_secs("apic"),
mod_name="ExpSyn",
netcon_weight=0.01,
delay=1,
number=100)
soma = cell.filter_secs("soma")
# Create IClamp
ic = IClamp(segment=soma(0.5))
ic.stim(delay=100, dur=10, amp=0.1)
# prepare plots and spike detector
rec_v = Record(soma(0.5), variables="v")
cell.make_spike_detector(soma(0.5))
# run
sim = Simulation(init_v=-65,
warmup=20,
init_sleep=2,
with_neuron_gui=True,
shape_plots=[make_shape_plot()])
sim.run(runtime=200, stepsize=1, delay_between_steps=500)
# plot
cell.plot_spikes()
rec_v.plot()
plt.show()
stim = NetStimCell("stim_cell").add_netstim(start=WARMUP + 1,
number=300,
interval=1)
cell.add_random_synapses_with_spine(source=stim,
secs=cell.secs,
mod_name="Syn4P",
netcon_weight=WEIGHT,
delay=1,
head_nseg=10,
neck_nseg=10,
number=10,
**cell.params_4p_syn)
# add mechanisms
cell.make_default_mechanisms()
cell.make_apical_mechanisms(sections='dend head neck')
# make plots
rec_w = Record(cell.filter_point_processes(mod_name="Syn4P",
name="head[0]"),
variables="w")
rec_v = Record(cell.filter_secs(name="head[0]")(1.0), variables="v")
# init and run
sim = Simulation(init_v=-70, warmup=WARMUP)
sim.run(runtime=500)
# plot
rec_w.plot()
rec_v.plot()
cell.group_synapses(tag="input_syn", synapses=[s, syn_ach, syn_da])
# add mechanisms
cell.make_default_mechanisms()
cell.make_apical_mechanisms(sections='dend head neck')
soma = cell.filter_secs("soma")
syns = cell.filter_synaptic_group()
syn4p = syns[0]['Syn4PAChDa']
synach = syns[0]['SynACh']
rec_syn = Record(syn4p, variables="w stdp_ach ach_stdp ACh ACh_w")
rec_soma = Record(soma(0.5), variables="v")
sim = Simulation(init_v=-80, warmup=WARMUP, warmup_on_create=True)
event = 0
inter = 5
for i in range(10):
for syn in syns:
syn['Syn4PAChDa'][0].make_event(event)
syn['SynACh'][0].make_event(event)
event += inter
sim.run(runtime=150)
# plot
rec_soma.plot()
rec_syn.plot()
spine_number=10,
spine_secs_names="apic",
spine_seed=13)
soma = cell.filter_secs("soma")
syns = cell.filter_complex_synapses(tag="combe")
# Prepare STDP protocol
stdp = Experiment()
stdp.make_protocol("3xEPSP[int=10] 3xAP[int=10,dur=3,amp=1.6]",
start=1,
isi=10,
epsp_synapse=syns[0],
i_clamp_section=soma)
# Prepare plots
v_soma_rec = Record([soma(0.5), syns[0].parent], variables='v')
cai_head0_rec = Record(syns[0].parent, variables='cai')
# Run
sim = Simulation(init_v=-70,
warmup=20,
with_neuron_gui=True,
constant_timestep=False)
sim.run(runtime=100, debug=True)
# Plot
cai_head0_rec.plot()
v_soma_rec.plot()
v_soma_rec.to_csv("vrec.csv")
soma = cell.filter_secs("soma")
# Netstim to synapse
stim = NetStimCell("stim").add_netstim(start=WARMUP,
number=REPS,
interval=interval)
syn = cell.add_synapse(source=stim,
netcon_weight=WEIGHT,
mod_name="Syn4P",
delay=1,
seg=cell.filter_secs('apic[1]')(0.5))
# IClamp to soma
iclamp = IClamp(segment=cell.filter_secs("soma")(0.5))
for i in range(REPS):
start_t = WARMUP + delta_t + i * interval
iclamp.stim(delay=start_t, dur=DUR, amp=AMP)
# Record
rec = Record([s(0.5) for s in cell.filter_secs("apic[1],apic[50]")])
# Run
sim = Simulation(init_v=-70, warmup=WARMUP, dt=DT)
total_time = REPS * interval + COOL_DOWN
sim.run(total_time)
# Plot
rec.plot(position="merge")
plt.show()
connector.build()
pop1.record()
# Create population 2
pop2 = Population("pop_2")
pop2.add_cells(num=4, cell_function=cell_function)
connector = pop2.connect(cell_connection_proba=connection_proba,
seg_dist=NormalTruncatedSegDist(0.5, 0.1))
connector.set_source([c.filter_secs("soma")(0.5) for c in pop1.cells])
connector.set_target([d(0.5) for c in pop1.cells for d in c.filter_secs("dend")])
syn_adder = connector.add_synapse("Exp2Syn")
syn_adder.add_netcon(weight=weight_dist)
connector.build()
pop2.record()
# Create connectivity graph grouped by populations, with weighs and spike rates updated
graph = NetworkGraph(populations=[pop1, pop2])
graph.plot()
# Run
sim = Simulation(init_v=-70, warmup=20)
for i in range(1000):
sim.run(runtime=1)
pop1.plot(animate=True)
pop2.plot(animate=True)
#graph.update_weights()
#graph.update_spikes()
class AgentCore:
def __init__(self, input_max_hz, default_stepsize):
"""
Before Agent step() you need to call:
1. agent.build()
2. agent.init()
:param input_max_hz:
:param default_stepsize:
"""
self.reward_syns = []
self.punish_syns = []
self.input_max_hz = input_max_hz
self.default_stepsize = default_stepsize
self.max_input_stim_per_stepsize = (default_stepsize *
input_max_hz) / 1000
if self.max_input_stim_per_stepsize < 1:
raise ValueError(
"Agent's self.max_input_stim_per_stepsize must be > 1, choose "
"input_max_hz and stepsize params carefully.")
print("max_input_stim_per_stepsize:", self.max_input_stim_per_stepsize)
self.sim = None
self.warmup = None
self.input_size = None
self.input_shape = None
self.input_cell_num = None
self.input_cells = None
self.output_cells = None
self._built = False
def init(self,
input_cells: List[Cell],
output_cells: List[Cell],
reward_syns: List[Synapse] = (),
punish_syns: List[Synapse] = (),
init_v=-70,
warmup=0,
dt=0.1):
"""
Before Agent step() you need to call:
1. agent.build()
2. agent.init()
:param input_cells:
list of input cells
:param output_cells:
list of output cells
:param init_v:
:param warmup:
:param dt:
"""
self.input_cells = input_cells
self.output_cells = output_cells
self.reward_syns = reward_syns
self.punish_syns = punish_syns
if self.sim is not None:
raise RuntimeError(
"Simulation cannot been run before initialization.")
self.warmup = warmup
self.sim = Simulation(init_v=init_v,
warmup=warmup,
dt=dt,
warmup_on_create=True)
def step(self,
observation: np.array,
output_type="time",
sort_func=None,
poisson=False,
stepsize=None):
"""
:param observation:
numpy array. 1 or 2 dim are allowed
:param output_type:
"time": returns time of first spike for each motor cells.
"rate": returns number of spikes for each motor cells OR -1 if there were no spike for
the cell.
"raw": returns raw array for each motor cell of all spikes in time in ms.
:param sort_func:
Optional function which define sorting on list of AgentOutput objects.
:param poisson:
if use Poisson distribution for each pixel stimulation. Default is False.
:param stepsize:
in ms. If None - it will use self.default_stepsize.
:return:
list(AgentOutput(index, cell_name, value))
"""
# Check agent's built and initialization before step
if not self._built:
raise RuntimeError(
"Before step you need to build() agent and then initialize by calling init() "
"function first.")
if self.sim is None:
raise RuntimeError(
"Before step you need to initialize the Agent by calling init() function first."
)
if self.input_cells is None or len(self.input_cells) == 0:
raise LookupError(
"Method self._build_network() must return tuple(input_cells, output_cells), "
"however input_cells were not defined.")
self._make_observation(observation=observation,
poisson=poisson,
stepsize=stepsize)
# Run
self._make_sim(stepsize)
# Make output
output = self._get_output(output_type)
if sort_func:
output = sorted(output, key=sort_func)
return output
def reward_step(self, reward, stepsize=None):
"""
It allows to sense the reward by the agent for stepsize time.
:param reward:
the value of the reward
:param stepsize:
in ms. If None - it will use self.default_stepsize.
"""
self._make_reward(reward)
self._make_sim(stepsize)
@staticmethod
def get_kernel_size(w, f, p, s):
"""
Naming convention comes from the Convolutional Neural Networks.
:param w:
image size of one of dimentions
:param f:
convolution size of one of dimentions
:param p:
padding
:param s:
stride
:return:
size of kernel for one of dimention
"""
return math.floor((w - f + 2 * p) / s + 1)
@abc.abstractmethod
def build(self, **kwargs):
raise NotImplementedError()
@abc.abstractmethod
def _make_observation(self, observation, poisson=False, stepsize=None):
raise NotImplementedError()
def _build(self, input_shape, input_size, input_cell_num):
"""
:param input_shape:
:param input_size:
:param input_cell_num:
"""
if self._built or self.sim is not None:
raise RuntimeError(
"You must first build agent before initialisation and run "
"the simulation.")
self.input_shape = input_shape
self.input_size = input_size
self.input_cell_num = input_cell_num
self._built = True
def _make_sim(self, stepsize=None):
if stepsize is None:
stepsize = self.default_stepsize
self.sim.run(stepsize)
def _make_reward(self, reward):
if not self._built:
raise RuntimeError(
"Before making reward you need to build the Agent by calling "
"build() function first.")
if self.sim is None:
raise RuntimeError(
"Before making reward you need to initialize the Agent by calling "
"init() function first.")
if reward > 0:
for s in self.reward_syns:
s.make_event(1)
elif reward < 0:
for s in self.punish_syns:
s.make_event(1)
def _get_output(self, output_type):
"""
:param output_type:
"time": returns time of first spike for each motor cells.
"rate": returns number of spikes for each motor cells OR -1 if there were no spike for
the cell.
"raw": returns raw array for each motor cell of all spikes in time in ms.
:return:
list(AgentOutput(index, value))
"""
outputs = []
min_time = self.sim.t - self.sim.current_runtime
for i, c in enumerate(self.output_cells):
spikes = np.array([i for i in c.spikes() if i >= min_time])
if output_type == "rate":
s = len(spikes) if len(spikes) > 0 else -1
elif output_type == "time":
s = spikes[0] if len(spikes) > 0 else -1
elif output_type == "raw":
s = spikes
else:
raise TypeError(
"Output type can be only string of: 'rate' or 'time', "
"but provided %s" % output_type)
outputs.append(AgentOutput(index=i, cell_name=c.name, value=s))
return outputs
def _make_single_observation(self,
observation,
syns,
poisson,
stepsize=None):
"""
The core observation method which match observation flat array (1D) to the list of synapses.
observation and syns need to be of the same length.
:param observation:
1 dim array of numbers
:param syns:
1 d array of synapses
"""
if len(observation) != len(syns):
raise ValueError(
"Single 1D observation or flatten kernel of 2D observation "
"must have the same length as provided subgroup of synapses.")
for pixel, syn in zip(observation, syns):
if pixel > 0:
stim_num, interval = self._get_stim_values(
pixel, poisson, stepsize)
next_event = 0
for e in range(stim_num):
syn.make_event(next_event)
next_event += interval
def _get_stim_values(self, pixel, poisson=False, stepsize=None):
"""
Returns number of events and their interval for a single synaptic stimulation
:param pixel:
single pixel value
:param poisson:
If use poisson distribution, Default is False.
:param stepsize:
stepsize for this observation. if default None - it will take self.default_stepsize
:return:
tuple(number of events, interval between events)
"""
stim_num = 0
stim_int = 0
if pixel <= 0:
return stim_num, stim_int
if stepsize is None:
stepsize = self.default_stepsize
max_stim = self.max_input_stim_per_stepsize
else:
max_stim = (stepsize * self.input_max_hz) / 1000
stim_num = int(round(pixel * max_stim))
if poisson:
stim_num = np.random.poisson(stim_num, 1)[0]
if stim_num > 0:
stim_int = stepsize / stim_num
return stim_num, stim_int
def _add_output_spike_detectors(self):
for oc in self.output_cells:
if not hasattr(oc, "_spike_detector"):
raise TypeError(
"Output cells must be of type NetConCell and have spike detection "
"mechanism.")
if oc._spike_detector is None:
soma = oc.filter_secs("soma")
if isinstance(soma, list):
raise LookupError(
"Output cells need to setup spike detector or at least have "
"a single 'soma' section so that spike detection can be "
"implemented automatically.")
oc.make_spike_detector(soma(0.5))
def _get_recursive_cells(self, obj):
acc = []
if isinstance(obj, CoreCell):
acc.append(obj)
elif isinstance(obj, Population):
acc.extend(obj.cells)
elif isinstance(obj, list):
for o in obj:
ac = self._get_recursive_cells(o)
acc.extend(ac)
return acc
def test_warmup(self):
rec = Record(self.soma(0.5))
value_error = False
try:
sim = Simulation(warmup=-100)
sim.run(100)
r1 = rec.as_numpy('v')
except ValueError:
value_error = True
self.assertTrue(value_error)
sim = Simulation(warmup=100, dt=1)
sim.run(100)
r2 = rec.as_numpy('v')
sim = Simulation(warmup=100, dt=1, warmup_dt=1)
sim.run(100)
r3 = rec.as_numpy('v')
sim.remove_immediate_from_neuron()
rec.remove_immediate_from_neuron()
self.assertEqual(111, r2.size)
self.assertEqual(201, r3.size)
def test_timestep_constant(self):
rec = Record(self.soma(0.5))
sim = Simulation(constant_timestep=True, dt=1)
sim.run(100)
r1 = rec.as_numpy('v')
sim = Simulation(constant_timestep=False, dt=1)
sim.run(100)
r2 = rec.as_numpy('v')
sim = Simulation(constant_timestep=False, dt=10)
sim.run(100)
r3 = rec.as_numpy('v')
sim = Simulation(constant_timestep=False, dt=0.0001)
sim.run(100)
r4 = rec.as_numpy('v')
sim.remove_immediate_from_neuron()
rec.remove_immediate_from_neuron()
self.assertEqual(101, r1.size)
self.assertEqual(188, r2.size)
self.assertEqual(180, r3.size)
self.assertEqual(189, r4.size)
from neuronpp.cells.graham2014_cell import Graham2014Cell from neuronpp.utils.simulation import Simulation cell = Graham2014Cell(name="cell") ampa_pp, nmda_pp = cell.filter_point_processes(parent="head[0]") sim = Simulation() sim.run(runtime=500)
cell.group_complex_synapses("input_syn", s, syn_ach, syn_da)
# add mechanisms
cell.make_default_mechanisms()
cell.make_apical_mechanisms(sections='dend head neck')
soma = cell.filter_secs("soma")
syns = cell.filter_complex_synapses()
syn4p = syns[0]['Syn4PAChDa']
synach = syns[0]['SynACh']
rec_syn = Record(syn4p, variables="w stdp_ach ach_stdp ACh ACh_w")
rec_soma = Record(soma(0.5), variables="v")
sim = Simulation(init_v=-80, warmup=WARMUP)
event = 0
inter = 5
for i in range(10):
for syn in syns:
syn['Syn4PAChDa'].make_event(event)
syn['SynACh'].make_event(event)
event += inter
sim.run(runtime=150)
# plot
rec_soma.plot()
rec_syn.plot()
from neuronpp.utils.record import Record
from neuronpp.utils.simulation import Simulation
path = os.path.dirname(os.path.abspath(__file__))
if __name__ == '__main__':
# Prepare cell
filepath = os.path.join(path, "..",
"commons/mods/sigma3syn")
cell = Cell("cell", compile_paths=filepath)
soma = cell.add_sec("soma", diam=20, l=20, nseg=10)
cell.insert('pas')
cell.insert('hh')
w = 0.003 # LTP
#w = 0.0022 # LTD
syn = cell.add_synapse(source=None, netcon_weight=w, seg=soma(0.5), mod_name="ExcSigma3Exp2Syn")
# prepare plots and spike detector
rec_v = Record(soma(0.5), variables="v")
rec_w = Record(syn, variables="w")
# run
sim = Simulation(init_v=-68, warmup=5)
syn.make_event(5)
sim.run(runtime=50)
# plot
rec_w.plot()
rec_v.plot()
# 2) Recommended synapse
syn1 = cell.add_synapse(source=stim,
seg=soma(0.5),
netcon_weight=0.01,
mod_name="Syn4P",
delay=1)
# 3) Event synapse
syn2 = cell.add_synapse(source=None,
seg=soma(0.5),
netcon_weight=0.01,
mod_name="Syn4P",
delay=1)
# prepare plots
rec_v = Record(soma(0.5), variables="v")
# run
sim = Simulation(init_v=-55, warmup=20, warmup_on_create=True)
# Making external events to the synapse
syn2.make_event(10)
syn2.make_event(20)
syn2.make_event(30)
syn2.make_event(40)
# Example of online update of the graph, however sim.run(runtime=1000) can be call in a single run
for i in range(1000):
sim.run(runtime=1)
rec_v.plot(animate=True)