nsims = 50 cost_fun = lambda s1, s2: (B1*s1-x)**2 + (z+B2*s2 - B2*(B1*s1-y)**2)**2 #perturb_rate = 0.01 # Proportion of points that are perturbations # # 1% = 10 Hz. Only half of these are spikes, so the injected noise rate is 5Hz mvals = [0.0025, 0.005, 0.01, 0.015] M = len(mvals) # Filename for results params_lif = ParamsLIF(sigma = sigma, tau = tau, mu = mu, c = c) lif = LIF(params_lif, t = tsim) lif.x = 0 t_filter = np.linspace(0, 1, 2000) exp_filter = np.exp(-t_filter/tau_s) exp_filter = exp_filter/np.sum(exp_filter) ds = exp_filter[0] wvals = np.linspace(wmin, wmax, Wn) beta_rd = np.zeros((Wn, Wn, n, T, nsims)) beta_rd_mean = np.zeros((Wn, Wn, n, T, nsims)) beta_rd_true = np.zeros((Wn, Wn, n, T, nsims)) beta_fd_true = np.zeros((Wn, Wn, n, T, nsims)) beta_sp = np.zeros((Wn, Wn, n, T, M, nsims))
x = .01
y = 0.1
z = 0
nsims = 50
cost_fun = lambda s1, s2: (B1*s1-x)**2 + (z+B2*s2 - B2*(B1*s1-y)**2)**2
#perturb_rate = 0.01 # Proportion of points that are perturbations
# # 1% = 10 Hz. Only half of these are spikes, so the injected noise rate is 5Hz
mvals = [0.0025, 0.005, 0.01, 0.015]
M = len(mvals)
params_lif = ParamsLIF(sigma = sigma, tau = tau, mu = mu, c = c)
lif = LIF(params_lif, t = tsim)
lif.x = x_input
t_filter = np.linspace(0, 1, 2000)
exp_filter = np.exp(-t_filter/tau_s)
exp_filter = exp_filter/np.sum(exp_filter)
ds = exp_filter[0]
wvals = np.linspace(wmin, wmax, Wn)
beta_rd = np.zeros((Wn, Wn, n, T, nsims))
beta_rd_mean = np.zeros((Wn, Wn, n, T, nsims))
beta_rd_true = np.zeros((Wn, Wn, n, T, nsims))
beta_fd_true = np.zeros((Wn, Wn, n, T, nsims))
beta_sp = np.zeros((Wn, Wn, n, T, M, nsims))
import numpy as np
from lib.lif import LIF, ParamsLIF
from lib.causal import causaleffect
#Set x = 0, sigma = 10
#wvals = 2..20
sigma = 10
mu = 1
tau = 1
t = 500
params = ParamsLIF(sigma = sigma, mu = mu, tau = tau)
lif = LIF(params, t = t)
lif.x = 0
#Simulate for a range of $W$ values.
N = 19
nsims = 1
wmax = 20
n = params.n
deltaT = 50
#Play with different c values
cvals = [0.01, 0.25, 0.5, 0.75, 0.99]
for c in cvals:
print("Running simulations for c = %f"%c)
outfile = './sweeps/param_w_N_%d_nsims_%d_c_%f_deltaT_counterfactual_simulations.npz'%(N, nsims, c)
params.c = c
lif.setup(params)
lif.x = 0
wvals = np.linspace(2, wmax, N)
cvals = np.array([0.01, 0.25, 0.5, 0.75, 0.99]) pvals = np.linspace(0.02, 1, 10) tau_s = 0.020 dt = 0.001 t = 50 alpha1 = -30 alpha2 = 20 x = 2 DeltaT = 50 params = ParamsLIF() params_orig = ParamsLIF() lif = LIF(params, t=t) t_filter = np.linspace(0, 0.150, 150) exp_filter = np.exp(-t_filter / tau_s) exp_filter = exp_filter / np.sum(exp_filter) ds = exp_filter[0] #c (correlation between noise inputs) beta_rd_c = np.zeros((len(cvals), nsims, len(pvals), params.n)) beta_fd_c = np.zeros((len(cvals), nsims, params.n)) beta_bp_c = np.zeros((len(cvals), nsims, params.n)) beta_rd_c_linear = np.zeros((len(cvals), nsims, len(pvals), params.n)) beta_fd_c_linear = np.zeros((len(cvals), nsims, params.n)) m_beta_bp_c = np.zeros(params.n)
cvals = np.array([0.01, 0.25, 0.5, 0.75, 0.99]) pvals = np.linspace(0.02, 1, 10) tau_s = 0.020 dt = 0.001 t = 50 alpha1 = -30 alpha2 = 20 x = 2 DeltaT = 50 params = ParamsLIF() params_orig = ParamsLIF() lif = LIF(params, t=t) t_filter = np.linspace(0, 0.150, 150) exp_filter = np.exp(-t_filter / tau_s) exp_filter = exp_filter / np.sum(exp_filter) ds = exp_filter[0] #c (correlation between noise inputs) beta_rd_c = np.zeros((len(cvals), nsims, len(pvals), params.n)) beta_fd_c = np.zeros((len(cvals), nsims, params.n)) beta_bp_c = np.zeros((len(cvals), nsims, params.n)) beta_rd_c_linear = np.zeros((len(cvals), nsims, len(pvals), params.n)) beta_fd_c_linear = np.zeros((len(cvals), nsims, params.n)) m_beta_bp_c = np.zeros(params.n)
import numpy as np
from lib.lif import LIF, ParamsLIF
from lib.causal import causaleffect
params = ParamsLIF()
lif = LIF(params)
#Simulate for a range of $W$ values.
N = 10
nsims = 500
wvals = np.linspace(1, 10, N)
beta_rd_w = np.zeros((N, N, nsims, params.n))
beta_fd_w = np.zeros((N, N, nsims, params.n))
beta_bp_w = np.zeros((N, N, nsims, params.n))
lif.setup(params)
p = 0.03
for i, w0 in enumerate(wvals):
for j, w1 in enumerate(wvals):
print("Running %d simulations with w0=%f, w1=%f" % (nsims, w0, w1))
lif.W = np.array([w0, w1])
for k in range(nsims):
(v, h, Cost, betas) = lif.simulate()
beta_rd_w[i, j, k, :] = causaleffect(v, Cost, p, params)
beta_fd_w[i, j, k, :] = causaleffect(v, Cost, 1, params)
beta_bp_w[i, j, k, :] = betas
#Save output
outfile = './sweeps/param_w_N_%d_nsims_%d_default.npz' % (N, nsims)
M = len(mvals)
# Filename for results
fn_out = './sweeps/learningbeta_fixedx_sweepw_banana_perturbation.npz'
#Cost function params
B1 = 2
B2 = 7
x = .05
y = 0.15
z = -0.2
params = ParamsLSM(q=q, p=1, t=t)
lsm = LSM(params)
params_lif = ParamsLIF()
lif = LIF(params_lif, t=t)
t_filter = np.linspace(0, 0.15, 150)
exp_filter = np.exp(-t_filter / tau_s)
exp_filter = exp_filter / np.sum(exp_filter)
ds = exp_filter[0]
wvals = np.linspace(1, wmax, Wn)
beta_rd = np.zeros((Wn, Wn, n, N))
beta_rd_true = np.zeros((Wn, Wn, n, N))
beta_fd_true = np.zeros((Wn, Wn, n, N))
beta_sp = np.zeros((Wn, Wn, n, N, M))
for i, w0 in enumerate(wvals):
print("W0=%d" % w0)
for j, w1 in enumerate(wvals):