P = float(args['-p'])
Q = float(args['-q'])
s = float(args['-s'])
if np.allclose(s, 0):
Ps = np.repeat(P, N)
Qs = np.repeat(Q, N)
else:
Ps = np.random.normal(P, P * s, size=N)
Qs = np.random.normal(Q, Q * s, size=N)
sigma = float(args['--sigma'])
# -
I, E = ie(t, Ps, Qs, N, dt=dt, sigma=sigma)
lfp = (E + I)
# -
save_kdf(
str(args['NAME']),
N=N,
E=E,
I=I,
lfp=lfp,
t=t,
dt=dt,
sigma=sigma,
Ps=Ps,
Qs=Qs)
if __name__ == "__main__":
args = docopt(__doc__, version='alpha')
# -
# Process params
t = float(args['-t'])
dt = float(args['--dt'])
P = float(args['-p'])
Q = float(args['-q'])
sigma = float(args['--sigma'])
# -
I, E = ie(t, P, Q, dt=dt, sigma=sigma)
lfp = E + I
# -
save_kdf(str(args['NAME']),
E=E,
I=I,
lfp=lfp,
t=t,
dt=dt,
sigma=sigma,
P=P,
Q=Q)
pass
np.random.seed(seed)
# -
# Process params
t = float(args['-t'])
dt = float(args['--dt'])
P0 = float(args['--p0'])
PN = float(args['--pn'])
Q = float(args['-q'])
sigma = float(args['--sigma'])
# -
# Run model
I, E = ie(t, P0, PN, sigma=sigma)
lfp = (E + I)
# -
save_kdf(str(args['NAME']),
E=E,
I=I,
lfp=lfp,
t=t,
dt=dt,
P0=P0,
PN=PN,
Q=Q,
sigma=sigma)
if M_no > (M_pre + (SD_pre * threshold)):
false_alarm = 1
hits[j, i] = hit
false_alarms[j, i] = false_alarm
# D prime
d_prime = (M_stim - M_no) / np.sqrt((SD_stim**2 + SD_no**2) / 2)
d_primes[j, i] = d_prime
# -------------------------------------------------------------------
# Calculate signal detection statistics
# First calculate misses and correct_rejections
misses = np.ones_like(hits) - hits
correct_rejections = np.ones_like(false_alarm) - false_alarms
# -------------------------------------------------------------------
# Save
save_kdf(
name,
hits=hits,
misses=misses,
correct_rejections=correct_rejections,
false_alarms=false_alarms,
d_primes=d_primes,
n_stim=n_stim,
stims=range(n_stim),
rates=rates)
M_right = M_pow2
p_left = np.exp(M_left) / np.sum([np.exp(M_left), np.exp(M_right)])
p_right = 1 - p_left
p_lefts[j, i] = p_left
p_rights[j, i] = p_right
# -------------------------------------------------------------------
# Calculate remaining signal detection statistics
misses = np.ones_like(hits) - hits
correct_rejections = np.ones_like(false_alarm) - false_alarms
# -------------------------------------------------------------------
# Save
save_kdf(
name,
hits=hits,
misses=misses,
correct_rejections=correct_rejections,
false_alarms=false_alarms,
p_lefts=p_lefts,
p_rights=p_rights,
d_primes=d_primes,
n_stim=n_stim,
stims=range(n_stim),
pow1=pow1,
powers2=powers2,
sigma=sigma,
loc=loc)
x_e = x_e[times > drop_before]
to_calc = ('HX', 'HY', 'HXY')
q_in, _, _ = en.quantise(x_in, m)
q_e, _, _ = en.quantise(x_e, m)
info = en.DiscreteSystem(q_in, (1, m), q_e, (1, m))
info.calculate_entropies(method='pt', calc=to_calc)
mi_pac = info.I()
# Save
del_mi[j, i] = mi_pac - mi_no
save_kdf(os.path.join(save_path, name),
del_mi=del_mi,
n_stim=n_stim,
n_sigma=n_sigma,
stims=range(n_stim),
sigmas=sigmas)
# -------------------------------------------------------------------
# b. osc
elif exp == "b":
name = "b"
nopac = os.path.join(pars_path, 'pac_osc_noalpha.yaml')
pac = os.path.join(pars_path, 'pac_osc.yaml')
del_mi = np.zeros((n_stim, n_sigma))
for i, s in enumerate(sigmas):
for j in range(n_stim):
print(">>> Fig {}, sigma {}, stim {}".format(name, s, j))
lif(
t,
Is,
f,
r_e=r_e * a,
r_i=r_i * a,
w_e=w_e,
w_i=w_i,
min_rate=min_rate * a, # TODO keep this scaling too?
n_bursts=n_bursts,
back_seed=42 * k,
verbose=False))
fis.append(np.vstack(fi_t).mean(0))
save_kdf(os.path.join(save_path, "a"), Is=Is, f=f, amps=amps, fis=fis)
# --
# b)
print("Running b)")
n_bursts = [2, 4, 6, 8]
fis = []
for n in n_bursts:
print("N bursts {}".format(n))
fi_t = []
for k in range(n_trials):
fi_t.append(
lif(t,
Is,
f,
# -
thetas, times = simulate(theta0, T, omegas, K, N, sigma, p, dt)
# -
# From the unit circle to sin waves, and the simulated lfp.
waves = []
for n in range(N):
th = thetas[:, n]
f = omegas[n]
wave = np.sin(f * 2 * np.pi * times + th)
waves.append(wave)
waves = np.vstack(waves)
lfp = waves.mean(0)
# -
save_kdf(str(args['NAME']),
thetas=thetas,
theta0=theta0,
omegas=omegas,
waves=waves,
lfp=lfp,
seed=seed,
sigma=sigma,
N=N,
K=K,
times=times,
t=T,
dt=dt)
r_i=r_i,
w_e=w_e,
w_i=w_i,
min_rate=min_rate,
return_trains=True)
ns2, ts2 = sp.spikedict_to(trains2)
save_kdf(
os.path.join(save_path, "a"),
Is=Is,
f=10,
fixed=1,
osc=2,
fi1=fi1,
fi2=fi2,
ns1=ns1,
ts1=ts1,
ns2=ns2,
ts2=ts2,
ge1=ge1,
gi1=gi1,
v1=v1,
ge2=ge2,
gi2=gi2,
v2=v2, )
# --
# b) phase plot, and full FI for peak/trough
# we turn osc off because were estimating gain at peak/trough by
# simulating at the peak/trough rates for longer periods of time than
# the peak/trough lasts peak.
f = 0