import numpy as np
import matplotlib.pyplot as plt
from ....Classes.OptiObjective import OptiObjective
from ....Methods.Output.XOutput import _get_symbol_data_
[docs]def plot_pareto(
self,
x_symbol,
y_symbol,
c_symbol=None,
cmap=None,
ax=None,
title=None,
grid=False,
is_show_fig=True,
save_path=None,
):
"""Plot the pareto front for 2 objective functions
Parameters
----------
self : OptiGenAlgNsga2Deap
x_symbol : str
symbol of the first objective function
y_symbol: str
symbol of the second objective function
c_symbol: str
optional symbol to set the plot colors
cmap: colormap
optional colormap
is_show_fig : bool
True to show figure after plot
save_path : str
full path of the png file where the figure is saved if save_path is not None
"""
# Pyleecan colors
pyleecan_color = (230 / 255, 175 / 255, 0)
# Gather fitness results
data = [
val.result
for _, val in self.xoutput.xoutput_dict.items()
if isinstance(val, OptiObjective)
]
fitness = np.array(data).T
# Get fitness values and ngen
is_valid = np.array(self.xoutput["is_valid"].result)
# ngen = np.array(self["ngen"].result) # unused
design_var_list = [pe.value for pe in self.xoutput.paramexplorer_list]
design_var = np.array(design_var_list).T
# Keep only valid values
indx = np.where(is_valid)
fitness = fitness[indx]
# ngen = ngen[indx] # unused
design_var = design_var[indx]
# get data and labels
x_values, x_label = _get_symbol_data_(self.xoutput, x_symbol, indx)
y_values, y_label = _get_symbol_data_(self.xoutput, y_symbol, indx)
# Get pareto front
pareto = fitness
# Get dominated values
idx_non_dom = list(range(len(pareto)))
N = len(pareto)
for i in range(N):
for j in idx_non_dom:
if all(pareto[j] <= pareto[i]) and any(pareto[j] < pareto[i]):
idx_non_dom.remove(i)
break
pareto = pareto[idx_non_dom]
design_var_values = design_var[idx_non_dom]
# Write annotations
legend_annot = []
design_var_symbols = [pe.symbol for pe in self.xoutput.paramexplorer_list]
for idx, sim in enumerate(design_var_values.tolist()):
legend = f"Individual Nr. {indx[0][idx_non_dom[idx]]}\n"
for ii, symbol in enumerate(design_var_symbols):
legend += "{:11.10}=".format(symbol) # sim[d_var])
if isinstance(sim[ii], float):
legend += " {:3.3e}\n".format(sim[ii])
else:
legend += "{:>11.10}\n".format(str(sim[ii]))
legend_annot.append(legend[:-1])
if ax is None:
fig, ax = plt.subplots()
return_ax = False
else:
return_ax = True
fig = ax.get_figure()
if c_symbol is None:
colors = pyleecan_color
else:
# get the color data
c_values, _ = _get_symbol_data_(self.xoutput, c_symbol, indx)
colors = c_values[idx_non_dom][:, np.newaxis]
if cmap is None:
cmap = plt.cm.jet
# Plot Pareto front
sc = ax.scatter(
x_values[idx_non_dom][:, np.newaxis],
y_values[idx_non_dom][:, np.newaxis],
# facecolors=colors,
c=colors,
edgecolors=(0.35, 0.35, 0.35),
label="Pareto Front",
cmap=cmap,
)
# Add legend
if c_symbol is not None:
legend1 = ax.legend(*sc.legend_elements(), loc="upper right", title=c_symbol)
ax.add_artist(legend1)
ax.autoscale(1, 1)
ax.set_title("Pareto Front")
ax.set_xlabel(x_label)
ax.set_ylabel(y_label)
if grid:
ax.set_axisbelow(True)
ax.grid()
# Add anotations in the plot see https://stackoverflow.com/a/47166787
annot = ax.annotate(
"",
xy=(0, 0),
xytext=(20, 20),
textcoords="offset points",
bbox=dict(boxstyle="round", fc="w"),
arrowprops=dict(arrowstyle="->"),
)
annot.set_visible(False)
def update_annot(ind):
"""Update annotation"""
# Get ind position
pos = sc.get_offsets()[ind["ind"][0]]
annot.xy = pos
# Set the annotation
annot.set_text(legend_annot[ind["ind"][0]])
annot.get_bbox_patch().set_facecolor(pyleecan_color)
annot.get_bbox_patch().set_alpha(0.4)
def hover(event):
# Check if annotation is visible
vis = annot.get_visible()
if event.inaxes == ax:
cont, ind = sc.contains(event)
if cont:
update_annot(ind)
annot.set_visible(True)
fig.canvas.draw_idle()
else:
if vis:
annot.set_visible(False)
fig.canvas.draw_idle()
fig.canvas.mpl_connect("motion_notify_event", hover)
if save_path is not None:
fig.savefig(save_path)
plt.close(fig=fig)
if is_show_fig:
fig.show()
if return_ax:
return ax
else:
fig.show()
