def step(self, y, predict_P=False, index=1):
X = self.x_prior
states = X[:2, :]
params = X[2:, :]
s_std = std(X[index].A1)
tmp_ws = as_array([norm.pdf(y, x[0, index], s_std) for x in states.T])
n_weights = self.weights * tmp_ws
sum_weights = n_weights.sum()
if sum_weights != 0:
n_weights /= sum_weights
neff = 1.0 / (n_weights**2).sum()
if sum_weights == 0 or neff < self.num_part / 2.0:
idx = choice(range(X.shape[1]), X.shape[1], p=self.weights)
self.weights = tile(as_array(1.0 / self.num_part), self.num_part)
self.x_post = X[:, idx]
else:
self.x_post = X
self.weights = n_weights
p_mean = average(params, axis=1, weights=self.weights).A1
p_cov = cov(params, aweights=self.weights)
self.x_post[2:, :] = multivariate_normal(p_mean, p_cov, X.shape[1]).T
for i, x in enumerate(self.x_post[2:, :].T):
if x.any() < 0:
while True:
new = multivariate_normal(p_mean, p_cov, 1).T
if new.all() > 0 and new[0, 1] > new[0, 2]:
self.x_post[2:, i] = new
break
def step(self, y, predict_P=False, index=1):
X = self.x_prior
states = X[:2, :]
params = X[2:, :]
s_std = std(X[index].A1)
tmp_ws = as_array([norm.pdf(y, x[0, index], s_std) for x in states.T])
n_weights = self.weights * tmp_ws
sum_weights = n_weights.sum()
if sum_weights != 0:
n_weights /= sum_weights
neff = 1.0 / (n_weights ** 2).sum()
if sum_weights == 0 or neff < self.num_part/2.0:
idx = choice(range(X.shape[1]), X.shape[1], p=self.weights)
self.weights = tile(as_array(1.0 / self.num_part), self.num_part)
self.x_post = X[:, idx]
else:
self.x_post = X
self.weights = n_weights
p_mean = average(params, axis=1, weights=self.weights).A1
p_cov = cov(params, aweights=self.weights)
self.x_post[2:, :] = multivariate_normal(p_mean, p_cov, X.shape[1]).T
for i, x in enumerate(self.x_post[2:, :].T):
if x.any() < 0:
while True:
new = multivariate_normal(p_mean, p_cov, 1).T
if new.all() > 0 and new[0, 1] > new[0, 2]:
self.x_post[2:, i] = new
break
def step(X, X_out, w, y, err, R, index=1):
s_std = std(X[index].A1)
tmp_ws = as_array([norm.pdf(y, x[0, index], s_std) for x in X.T])
# It is strange that the numerical instability was never found in
# the python 2.7 version.
# The following two lines are just an ad-hoc solution
if (tmp_ws == 0).all():
tmp_ws = w
else:
tmp_ws *= w
tmp_ws /= tmp_ws.sum()
s_idx = where(tmp_ws >= err)[0]
l_idx = where(tmp_ws < err)[0]
if l_idx.shape[0] > 0:
s_ws = tmp_ws[s_idx] / tmp_ws[s_idx].sum()
noise = multivariate_normal(zeros(X.shape[0]), R, l_idx.shape[0])
idx = choice(s_idx, l_idx.shape[0], p=s_ws)
X_out[:, l_idx] = X_out[:, idx] + as_matrix(noise).T
tmp_ws[l_idx] = tmp_ws[idx]
w[:] = tmp_ws / tmp_ws.sum()
else:
w[:] = tmp_ws[:]
return X_out, w, l_idx.shape[0]
def step(X, X_out, w, y, err, R, index=1):
s_std = std(X[index].A1)
tmp_ws = as_array([norm.pdf(y, x[0, index], s_std) for x in X.T])
# It is strange that the numerical instability was never found in
# the python 2.7 version.
# The following two lines are just an ad-hoc solution
if (tmp_ws == 0).all():
tmp_ws = w
else:
tmp_ws *= w
tmp_ws /= tmp_ws.sum()
s_idx = where(tmp_ws >= err)[0]
l_idx = where(tmp_ws < err)[0]
if l_idx.shape[0] > 0:
s_ws = tmp_ws[s_idx] / tmp_ws[s_idx].sum()
noise = multivariate_normal(zeros(X.shape[0]), R, l_idx.shape[0])
idx = choice(s_idx, l_idx.shape[0], p=s_ws)
X_out[:, l_idx] = X_out[:, idx] + as_matrix(noise).T
tmp_ws[l_idx] = tmp_ws[idx]
w[:] = tmp_ws / tmp_ws.sum()
else:
w[:] = tmp_ws[:]
return X_out, w, l_idx.shape[0]
def step(self, y, predict_P=False, index=1):
X = self.x_prior
states = X[:2, :]
params = X[2:, :]
print('A1', X[index].A1.shape)
s_std = std(X[index].A1)
tmp_ws = as_array([norm.pdf(y, x[0, index], s_std) for x in states.T])
n_weights = self.weights * tmp_ws
sum_weights = n_weights.sum()
if sum_weights != 0:
n_weights /= sum_weights
neff = 1.0 / (n_weights**2).sum()
if sum_weights == 0 or neff < self.num_part / 2.0:
print('index', index, 'resamples')
idx = choice(range(X.shape[1]), X.shape[1], p=self.weights)
self.weights = tile(as_array(1.0 / self.num_part), self.num_part)
self.x_post = X[:, idx]
else:
self.x_post = X
self.weights = n_weights
print('weight:', self.weights.shape, 'n_weight', n_weights.sum(),
'curr weight', self.weights.sum())
p_mean = average(params, axis=1, weights=self.weights).A1
p_cov = cov(params, aweights=self.weights)
self.x_post[2:, :] = multivariate_normal(p_mean, p_cov, X.shape[1]).T
for i, x in enumerate(self.x_post[
2:, :].T): #iterate over all parameter post estimat.
# resample any particle parameters (alpha,beta)
if x.any() < 0:
while True:
new = multivariate_normal(p_mean, p_cov, 1).T
if new.all() > 0 and new[0, 1] > new[0, 2]:
self.x_post[2:, i] = new
break
def get_score(self):
I_mat = as_array(self.Is)
for i, w in enumerate(self.weights):
I_mat[i] *= w
self.IS = sum(I_mat, axis=1)
time_gap = 1 # self.epochs / 52
idx = [x for x in range(self.epochs) if not x % time_gap]
self.score = RSS(self.CDC_obs, self.IS[idx])
self.scores = {}
self.scores['SSE'] = self.score
self.scores['RMSE'] = RMSE(self.CDC_obs, self.IS[idx])
self.scores['MSPE'] = MSPE(self.CDC_obs, self.IS[idx])
self.scores['CORR'] = corrcoef(self.CDC_obs, self.IS[idx])[0, 1]
return self.score
def get_score(self):
I_mat = as_array(self.Is)
for i, w in enumerate(self.weights):
I_mat[i] *= w
self.IS = sum(I_mat, axis=1)
time_gap = self.epochs / 52
idx = [x for x in range(self.epochs) if not x % time_gap]
self.score = RSS(self.CDC_obs, self.IS[idx])
self.scores = {}
self.scores['SSE'] = self.score
self.scores['RMSE'] = RMSE(self.CDC_obs, self.IS[idx])
self.scores['MSPE'] = MSPE(self.CDC_obs, self.IS[idx])
self.scores['CORR'] = corrcoef(self.CDC_obs, self.IS[idx])[0, 1]
return self.score
