theta_true = 0.1
Qxi = Q[0, 0].reshape((1, 1))
Qz = Q[1, 1].reshape((1, 1))
P0 = numpy.eye(2)
# How many steps forward in time should our simulation run
steps = 200
# model = ParticleParamTrans((theta_true,), R=R, Qxi=Qxi, Qz=Qz)
model = ParticleParamTrans((theta_true, ), R=R, Qxi=Qxi, Qz=Qz)
if (gradient_test):
num = 50
nums = 5
# numpy.random.seed(1)
x0 = numpy.vstack((xi0_true, z0_true))
(y, x) = generate_reference(x0, P0, theta_true, steps)
gt = gradienttest.GradientTest(model, u=None, y=y)
gt.set_params(numpy.array((theta_true, )))
gt.simulate(num, nums)
param_steps = 101
param_vals = numpy.linspace(-0.1, 0.4, param_steps)
gt.test(0, param_vals, num=num, nums=nums)
plt.ion()
plt.figure(1)
plt.clf()
plt.plot(range(steps + 1), x[:, 0], 'r-')
plt.plot(range(steps + 1), x[:, 1], 'b-')
sest = gt.straj.get_smoothed_estimates()
for j in xrange(nums):
plt.plot(range(steps + 1), sest[:, j, 0], 'g--')
plt.plot(range(steps + 1), sest[:, j, 1], 'k--')
if __name__ == '__main__':
theta_true = 2.5
# How many steps forward in time should our simulation run
steps = 200
uvec = -10.0 * numpy.hstack(
(-1.0 * numpy.ones(steps / 4), 1.0 * numpy.ones(steps / 2),
-1.0 * numpy.ones(steps / 4)))
if (gradient_test):
num = 50
nums = 5
# numpy.random.seed(1)
model = ParticleParamOutput((theta_true, ))
(u, y, x) = generate_reference(uvec, steps, theta_true)
gt = gradienttest.GradientTest(model, u=u, y=y)
gt.set_params(numpy.array((theta_true, )))
gt.simulate(num, nums)
param_steps = 101
param_vals = numpy.linspace(-10.0, 10.0, param_steps)
gt.test(0, param_vals, num=num, nums=nums)
plt.ion()
plt.figure(1)
plt.clf()
plt.plot(range(steps + 1), x[:, 0], 'r-')
plt.plot(range(steps + 1), x[:, 1], 'b-')
sest = gt.straj.get_smoothed_estimates()
for j in xrange(nums):
plt.plot(range(steps + 1), sest[:, j, 0], 'g--')
plt.plot(range(steps + 1), sest[:, j, 1], 'k--')