def test_build_stimulus_currents():
stim = ABAStimulus()
# center, spread for each unit, starting at 440 Hz A and 3 semitones up from there
s_units = [
Selectivity(music.key_to_frequency(49), 1, 0.6),
Selectivity(music.key_to_frequency(46), 1, 0.6)
]
stim = ABAStimulus(a_semitone=49, df=3)
network = SynapticNetwork(selectivities=s_units, stimulus=stim)
out = network.build_stimulus_currents()
assert out[0][0] > out[1][0]
assert out[0][50] < out[1][50]
def test_tones_init():
stim = ABAStimulus()
# print(stim.__dict__)
assert len(stim.tax) == len(stim.tones) == stim.ttot
assert stim.tones[0] == 49
assert stim.tax[-1] == stim.T # tax[0] is not 0...
assert not any([np.isnan(x) for x in stim.tones])
def test_triplets_timing():
tc = OrderedDict(
[
(43.0, 0.1353352832366127), (44.0, 0.24935220877729614), (45.0, 0.41111229050718745),
(46.0, 0.6065306597126333), (47.0, 0.8007374029168081), (48.0, 0.9459594689067655), (49.0, 1.0),
(50.0, 0.9459594689067655), (51.0, 0.8007374029168081), (52.0, 0.6065306597126333),
(53.0, 0.41111229050718745),
(54.0, 0.24935220877729614)
]
)
#df, iti, tone_length, dt = 4, .04, .033, .001
#actual = aba_triplet(tc, df, iti=iti, tone_length=tone_length, dt=dt)
stim = ABAStimulus()
actual = stim.generate_triplet_tones()
# we wanna check the tone length... of all 3...
tone_time = int(stim.tone_length/stim.dt)
iti_time = int(stim.iti/stim.dt)
gap_time = iti_time - tone_time
# for u_ind in xrange(actual.shape[0]):
for t_ind in xrange(3):
start_ind = iti_time * t_ind
# check the tone length
assert list(actual[start_ind:]).index(0) == tone_time
# check the gap length
if t_ind is not 2:
assert list(actual[start_ind + tone_time:] > 0).index(True) == gap_time
# check the period
assert len(actual) == iti_time * 4
assert not any([np.isnan(x) for x in actual])
def test_map_tones_to_cell():
u1 = SensoryWCUnit()
stimulus = ABAStimulus()
u1.add_stim_current(stimulus, .5)
# print(u1.currents)
# assert False
assert u1.stim[0] == 0
u1.currents["stim"].update()
assert u1.currents["stim"].value == 1
def test_selectivity():
Selectivity = namedtuple("Selectivity",
("best_frequency", "spread", "gain"))
# center, spread for each unit, starting at 440 Hz A and 3 semitones up from there
s_units = [
Selectivity(music.key_to_frequency(49), 1, 0.8),
Selectivity(music.key_to_frequency(52), 1, 0.8)
]
stim = ABAStimulus()
network = TonotopicNetwork(s_units, stim)
network.update_all(10)
assert network.units[0].r > network.units[1].r
def test_pars_list():
Selectivity = namedtuple("Selectivity",
("best_frequency", "spread", "gain"))
pars_list = [{"tau": 100., "the": .4}, None]
# center, spread for each unit, starting at 440 Hz A and 3 semitones up from there
s_units = [
Selectivity(music.key_to_frequency(49), 1.1, 0.8),
Selectivity(music.key_to_frequency(49), 1, 0.8)
]
stim = ABAStimulus()
network = TonotopicNetwork(s_units, stim, pars_list=pars_list)
network.update_all(10)
# print(network.units[0].__dict__)
# print(network.units[1].__dict__)
assert network.units[0].r < network.units[1].r
# update response
self.unit.update()
# update traces
for trace in self.traces.values():
trace.update_trace()
self.t_i += 1
@staticmethod
def main():
pass
if __name__ == '__main__':
tic = datetime.datetime.now()
u1 = SensoryWCUnit(name="u1", tauA=5000, gSFA=0.8)
stim = ABAStimulus(a_semitone=44, df=9)
u1.add_stim_current(stim, weight=0.5)
sim = SensoryTripletsSimulation(sensory_unit=u1, T=5)
sim.run()
toc = datetime.datetime.now()
# let's try it with seaborn and a dataframe
trace_dict = {}
trace_dict.update((k, v.trace) for k, v in sim.traces.items())
trace_dict['tax'] = sim.tax
data = pd.DataFrame(trace_dict, index=sim.tax) \
[['tax', 'FR', 'SFA', 'stim']]
plot_sensory_traces(data=data, unit=u1)
# data1 = data.add(pd.Series(np.ones(len(sim.tax)), index=sim.tax))
df["tax"] = sim.tax
ulst = [df]
for k, v in units.items()[1:]:
ndf = pd.DataFrame(v) # units.items()[0][1], index=sim.tax)
ndf["unit"] = k # units.items()[0][0]
ndf["tax"] = sim.tax
ulst.append(ndf)
df_out = pd.concat(ulst)
return df_out
if __name__ == '__main__':
tic = datetime.datetime.now()
stim = ABAStimulus()
Selectivity = namedtuple("Selectivity", ("best_frequency", "spread", "gain"))
# center, spread for each unit, starting at 440 Hz A and 3 semitones up from there
s_units = [
Selectivity(music.key_to_frequency(49), 1, 0.6),
Selectivity(music.key_to_frequency(52), 1, 0.6)
]
stim = ABAStimulus(a_semitone=49, df=3)
network = TonotopicNetwork(s_units, stim)
sim = TonotopicTripletsSimulation(network=network)
sim.run()
data = sim.traces_to_df()
g = sns.FacetGrid(data, col='unit', col_wrap=1)
g.map_dataframe(plot_generic_traces)
"""
Add a unit to the network.
Args:
selectivity: tuple of (best_frequency, spread, gain)
best_frequency (float): Hz, e. g., 440.0
spread (float): by default, standard deviation of norm response curve in semitones away from BF
gain (float): strength of the stimulus on the cell.
"""
best_frequency, spread, gain = selectivity
name = name = 's({st}:{sp}:{g})'.format(st=int(
music.frequency_to_key(best_frequency)),
sp=spread,
g=gain)
unit = SensoryWCUnit(best_frequency=best_frequency,
spread=spread,
name=name,
**kwargs)
unit.add_stim_current(self.stimulus, weight=gain)
self.units.append(unit)
if __name__ == "__main__":
tic = time.time()
stim = ABAStimulus()
network = TonotopicNetwork(s_units, stim)
network.update_all(10)
toc = time.time()
print(toc - tic)
print(network.units[0].r, network.units[1].r)
print(stim.value)