def simulate_with_residuals(self, residuals, ndisc = 100):
"""Simulate a time series using this model using the supplied residuals
in residuals (ndarray). This function is NOT deterministic and employs
a spin-up phase to init the model state using random noise."""
A = self.A
w = self.w
m, p = self.dimension(), self.order()
N = residuals.shape[0]
ts = np.zeros((N, m))
# predictors always start with zeros
u = np.zeros((m*p,), dtype=np.float64)
# initialize system using noise with correct covariance matrix if required
if ndisc > 0:
eps_noise = np.dot(np.random.normal(size=(ndisc, m)), self.U)
# spin-up to random state, which is captured by vector u
var_model_acc.execute_Aupw(A, w, u, eps_noise, eps_noise)
# start feeding in residuals and store result in ts, start with vector u as VAR(p) state
var_model_acc.execute_Aupw(A, w, u, residuals, ts)
return ts
def simulate_with_residuals(self, residuals, ndisc=100):
"""Simulate a time series using this model using the supplied residuals
in residuals (ndarray). This function is NOT deterministic and employs
a spin-up phase to init the model state using random noise."""
A = self.A
w = self.w
m, p = self.dimension(), self.order()
N = residuals.shape[0]
ts = np.zeros((N, m))
# predictors always start with zeros
u = np.zeros((m * p, ), dtype=np.float64)
# initialize system using noise with correct covariance matrix if required
if ndisc > 0:
eps_noise = np.dot(np.random.normal(size=(ndisc, m)), self.U)
# spin-up to random state, which is captured by vector u
var_model_acc.execute_Aupw(A, w, u, eps_noise, eps_noise)
# start feeding in residuals and store result in ts, start with vector u as VAR(p) state
var_model_acc.execute_Aupw(A, w, u, residuals, ts)
return ts
def simulate(self, N, ndisc = 100):
"""Simulate process described by the model. Obtain N samples and spin up the model for 100 steps."""
A = self.A
w = self.w
m, p = A.shape[0], A.shape[1] / A.shape[0]
ts = np.zeros((N, m))
u = np.zeros((m*p,))
# construct noise with covariance matrix L * L^T
eps_noise = np.dot(np.random.normal(size=(N + ndisc, m)), self.U)
# spin up the model by running it for ndisc samples
var_model_acc.execute_Aupw(A, w, u, eps_noise[:ndisc, :], eps_noise[:ndisc, :])
# generate requested number of points
var_model_acc.execute_Aupw(A, w, u, eps_noise[ndisc:, :], ts)
return ts
def simulate(self, N, ndisc = 100):
"""Simulate process described by the model. Obtain N samples and spin up the model for 100 steps."""
A = self.A
w = self.w
m, p = A.shape[0], A.shape[1] / A.shape[0]
ts = np.zeros((N, m))
u = np.zeros((m*p,))
# construct noise with covariance matrix L * L^T
eps_noise = np.dot(np.random.normal(size=(N + ndisc, m)), self.U)
# spin up the model by running it for ndisc samples
var_model_acc.execute_Aupw(A, w, u, eps_noise[:ndisc, :], eps_noise[:ndisc, :])
# generate requested number of points
var_model_acc.execute_Aupw(A, w, u, eps_noise[ndisc:, :], ts)
return ts
