def gmrf_est(gmrf, process, experiment_name=None, gmrf_estimation_parameters=None):
''' Saves and creates a gmfr estimator from a GMRF.
Arguments:
gmrf: a GMRF object
process: the name of the process (diagonal, jl, exact)
experiment_name -- (string), the name of the experiment
gmrf_estimation_parameters -- (dict) the parameters, if any. depends on process
Returns:
a GMRFEstimator object
Side effects:
May save some information, if experiment_name is providide
'''
if process == 'diagonal':
diag_variance = 1.0 / gmrf.diag_precision
gmrf_estimator = DiagonalGMRFEstimator(gmrf.translations, gmrf.means, diag_variance)
if experiment_name is not None:
save_gmrf_estimator_DiagonalGMRFEstimator(gmrf_estimator, experiment_name=experiment_name)
elif process == 'jl':
precision = gmrf.precision
k = gmrf_estimation_parameters['k']
Q = random_projection_cholmod_csc(precision, k)
print "Q shape",Q.shape
gmrf_estimator = JLGMRFEstimator(gmrf.translations, gmrf.means, Q)
if experiment_name is not None:
save_gmrf_estimator_JLGMRFEstimator(gmrf_estimator, experiment_name=experiment_name)
elif process == 'exact':
X = gmrf.precision.todense()
covariance = np.linalg.inv(X)
gmrf_estimator = ExactGMRFEstimator(gmrf.translations, gmrf.means, covariance)
if experiment_name is not None:
save_gmrf_estimator_ExactGMRFEstimator(gmrf_estimator, experiment_name=experiment_name)
else:
assert False
return gmrf_estimator
return (D,P,rows,cols) (D,P,rows,cols) = star(5,4) X = build_dense(D, P, rows, cols) Xs = build_sparse(D, P, rows, cols) l1 = logdet_dense(D, P, rows, cols) l2 = logdet_dense_chol(D, P, rows, cols) l3 = logdet_cholmod(D, P, rows, cols) (M,Dn,Pn) = normalized_problem(D, P, rows, cols) test_data(D, P, rows, cols) W = la.inv(X) #Q = random_projection_cholmod(D, U, rows, cols, k, factor) Q = random_projection_cholmod_csc(Xs, k=1000) A = Q.T print A.shape R = np.sum(A*A,axis=1) U = np.sum(A[rows]*A[cols],axis=1) R_ = W.diagonal() U_ = W[rows,cols] #X = build_sparse(D, P, rows, cols) #(eis,_)=eigsh(X, k=1, sigma=-1, which='LM') #ei = eis[0] #is_psd_dense(R, U, rows, cols) #is_psd_cholmod(R, U, rows, cols) (R,U) = inv_dense(D, P, rows, cols)
