def c0_all(trajs):
nfeatures = get_nfeatures(trajs)
numer = np.zeros((nfeatures, nfeatures))
denom = 0.0
for traj in trajs:
numer += traj.T @ traj
denom += len(traj)
return numer / denom
def c0_trunc(trajs, cutlag):
nfeatures = get_nfeatures(trajs)
numer = np.zeros((nfeatures, nfeatures))
denom = 0.0
for traj in trajs:
x = traj[: len(traj) - cutlag]
numer += x.T @ x
denom += len(x)
return numer / denom
def c0_rt(trajs, cutlag, weights):
nfeatures = get_nfeatures(trajs)
numer = np.zeros((nfeatures, nfeatures))
denom = 0.0
for traj, weight in zip(trajs, weights):
x = traj[: len(traj) - cutlag]
w = weight[: len(traj) - cutlag]
numer += np.einsum("n,ni,nj", w, x, x)
denom += np.sum(w)
return numer / denom
def c0_all_adj_ct(trajs, lag):
nfeatures = get_nfeatures(trajs)
numer = np.zeros((nfeatures, nfeatures))
denom = 0.0
for traj in trajs:
x = traj[: len(traj) - lag]
y = traj[lag:]
numer += x.T @ x + y.T @ y
denom += 2 * len(x)
return numer / denom
def ct_ra(trajs, lag, cutlag, weights):
nfeatures = get_nfeatures(trajs)
numer = np.zeros((nfeatures, nfeatures))
denom = 0.0
for traj, weight in zip(trajs, weights):
w = weight[: len(traj) - cutlag]
for shift in range(cutlag - lag + 1):
x = traj[shift : shift + len(traj) - cutlag]
y = traj[shift + lag : shift + len(traj) - cutlag + lag]
numer += np.einsum("n,ni,nj", w, x, y)
denom += np.sum(w)
return numer / denom
def ic_trunc(trajs, lags, cutlag):
nfeatures = get_nfeatures(trajs)
ic = np.zeros((nfeatures, nfeatures))
for lag in lags:
ic += ct_trunc(trajs, lag, cutlag)
return ic
def c0_all_adj_ic(trajs, lags):
nfeatures = get_nfeatures(trajs)
c0 = np.zeros((nfeatures, nfeatures))
for lag in lags:
c0 += c0_all_adj_ct(trajs, lag)
return c0 / len(lags)
def ic_all(trajs, lags):
nfeatures = get_nfeatures(trajs)
ic = np.zeros((nfeatures, nfeatures))
for lag in lags:
ic += ct_all(trajs, lag)
return ic
def c0_ra_adj_ic(trajs, lags, cutlag, weights):
nfeatures = get_nfeatures(trajs)
c0 = np.zeros((nfeatures, nfeatures))
for lag in lags:
c0 += c0_ra_adj_ct(trajs, lags, cutlag, weights)
return c0 / len(lags)
def ic_ra(trajs, lags, cutlag, weights):
nfeatures = get_nfeatures(trajs)
ic = np.zeros((nfeatures, nfeatures))
for lag in lags:
ic += ct_ra(trajs, lag, cutlag, weights)
return ic
