landweber = Tikhonov(
kernel=kernel_transformed,
singular_values=spectrum.singular_values,
left_singular_functions=spectrum.left_functions,
right_singular_functions=spectrum.right_functions,
observations=obs,
sample_size=s,
max_size=max_size,
tau=tau,
order=order,
transformed_measure=True,
njobs=-1)
landweber.estimate()
landweber.oracle(fun)
solution = list(
landweber.solution(np.linspace(0, 1, 10000)))
results['selected_param'].append(
landweber.regularization_param)
results['oracle_param'].append(landweber.oracle_param)
results['oracle_loss'].append(landweber.oracle_loss)
results['loss'].append(landweber.residual)
results['solution'].append(solution)
results['oracle_solution'].append(
list(landweber.oracle_solution))
landweber.client.close()
except:
pass
pd.DataFrame(results).to_csv(
'Tikhonov1times_{}_{}_tau_{}.csv'.format(
functions_name[i], s, taus_name[j]))
kernel=kernel,
singular_values=spectrum.singular_values,
left_singular_functions=spectrum.
left_functions,
right_singular_functions=spectrum.
right_functions,
observations=obs,
sample_size=s,
transformed_measure=True,
max_size=max_size,
order=order,
njobs=-1)
tikhonov.estimate()
tikhonov.oracle(fun, patience=50)
solution = list(
tikhonov.solution(np.linspace(0, 1, 10000)))
results['selected_param'].append(
tikhonov.regularization_param)
results['oracle_param'].append(
tikhonov.oracle_param)
results['oracle_loss'].append(tikhonov.oracle_loss)
results['loss'].append(tikhonov.residual)
results['solution'].append(solution)
results['oracle_solution'].append(
list(tikhonov.oracle_solution))
tikhonov.client.close()
except:
pass
pd.DataFrame(results).to_csv(
'Test1_Tikhonov_{}_{}_{}_{}.csv'.format(
functions_name[i], s, order, taus_name[j]))