import os
from pathlib import Path
import matplotlib.pyplot as plt
import numpy as np
import numpy.linalg as npl
[docs]
def plot_mean_error_ctrl_comparison_vs_ite(
case_dict,
ctrl_target,
optim_iter_start=0,
optim_iter_end=-1,
flag_domain=None,
ordre_norm=2,
ylabel=None,
save_path=None,
file_name=None,
file_format='pdf',
figsize=(20, 15),
):
"""Plot mean error with control, comparison with different cases
Args:
case_dict (dict of cases): the different cases
ctrl_target (dataframe): control target
optim_iter_start (int, optional): iter start. Defaults to 0.
optim_iter_end (int, optional): iter end. Defaults to -1.
flag_domain (_type_, optional): flag of the domain. Defaults to None.
ordre_norm (int, optional): error norm order. Defaults to 2.
ylabel (str, optional): ylabel. Defaults to None.
save_path (str, optional): save path. Defaults to None.
file_name (str, optional): file name. Defaults to None.
file_format (str, optional): file format. Defaults to 'pdf'.
figsize (tuple, optional): figure size. Defaults to (20, 15).
"""
if flag_domain is None:
flag_domain = np.full(ctrl_target.value.shape, True, dtype=bool)
if ylabel is None:
ylabel = r'$RMSE(s)=\sqrt{\frac{1}{|\Omega|T}\int_\Omega\int_0^T(s-s_t)^2dtdxdy}~~(m^2.s^{-1})$'
fontsize = 25
plt.figure(figsize=figsize)
for key in case_dict.keys():
case = case_dict[key]
mean_error = [
npl.norm(s[flag_domain] - ctrl_target.value[flag_domain], ord=ordre_norm) / (len(s[flag_domain]) ** (1 / ordre_norm))
for s in case['S vs ite']
]
plt.plot(mean_error[optim_iter_start:optim_iter_end], case['ls'], label=case['label'])
plt.xlabel(r'iteration', fontsize=fontsize)
plt.ylabel(ylabel, fontsize=fontsize)
plt.tick_params(labelsize=fontsize - 5)
plt.yscale('log')
plt.legend(fontsize=fontsize)
if save_path is None:
plt.show()
else:
# if the save directory does not exist, then it is created
if not os.path.isdir(Path(save_path)):
os.makedirs(Path(save_path))
if file_name is None:
file_name = f'mean_error_ordre_{ordre_norm}_comparison'
plt.savefig(Path(save_path) / f"{file_name}.{file_format}", format=file_format)
plt.close('all')