def etd_solution(P, G, L, X, ivec, r):
# compute intermediate quantities (could be more efficient)
di = mdputils.stationary(P) * ivec
m = mult(resolvent(L, G, P.T), potential(G, P.T), di)
M = np.diag(m)
# solve the equation
A = mult(X.T, M, potential(P, G, L), resolvent(P, G), X)
A_inv = np.linalg.pinv(A)
b = mult(X.T, M, potential(P, G, L), r)
return np.array(np.dot(A_inv, b))
def etd_solution(P, G, L, X, ivec, r):
# compute intermediate quantities (could be more efficient)
di = mdputils.stationary(P) * ivec
m = mult(resolvent(L, G, P.T), potential(G, P.T), di)
M = np.diag(m)
# solve the equation
A = mult(X.T, M, potential(P, G, L), resolvent(P, G), X)
A_inv = np.linalg.pinv(A)
b = mult(X.T, M, potential(P, G, L), r)
return np.array(np.dot(A_inv, b))
def td_solution(P, G, L, X, r):
D = mdputils.distribution_matrix(P)
A = mult(X.T, D, resolvent(P, G), X)
A_inv = np.linalg.pinv(A)
b = mult(X.T, D, r)
return np.array(np.dot(A_inv, b))
def least_squares(P, G, X, r):
"""Compute the optimal weights via least squares."""
v = bellman(P, G, r)
D = mdputils.distribution_matrix(P)
return np.array(mult(np.linalg.pinv(mult(X.T, D, X)), X.T, D, v))
def td_solution(P, G, L, X, r):
D = mdputils.distribution_matrix(P)
A = mult(X.T, D, resolvent(P, G), X)
A_inv = np.linalg.pinv(A)
b = mult(X.T, D, r)
return np.array(np.dot(A_inv, b))
def least_squares(P, G, X, r):
"""Compute the optimal weights via least squares."""
v = bellman(P, G, r)
D = mdputils.distribution_matrix(P)
return np.array(mult(np.linalg.pinv(mult(X.T, D, X)), X.T, D, v))
