Skip to content

Commit

Permalink
place legends on top, increase font size
Browse files Browse the repository at this point in the history
  • Loading branch information
@bheijden
bheijden committed Sep 5, 2023
1 parent 15f1482 commit ff4ec57
Showing 1 changed file with 130 additions and 12 deletions.
142 changes: 130 additions & 12 deletions scripts/evaluate_graphs_plots.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -10,6 +10,17 @@
import supergraph.open_colors as oc


def export_legend(fig, legend, expand=None):
expand = [-5, -5, 5, 5] if expand is None else expand
# fig = legend.figure
# fig.canvas.draw()
bbox = legend.get_window_extent()
bbox = bbox.from_extents(*(bbox.extents + np.array(expand)))
bbox = bbox.transformed(fig.dpi_scale_trans.inverted())
return bbox
# fig.savefig(filename, dpi="figure", bbox_inches=bbox)


if __name__ == "__main__":
# Setup sns plotting
import seaborn as sns
Expand All @@ -23,11 +34,11 @@
scaling = 5
MUST_BREAK = False
plt.rcParams['axes.labelweight'] = 'bold'
plt.rcParams['axes.labelsize'] = 5 * scaling
plt.rcParams['legend.fontsize'] = 4 * scaling
plt.rcParams['font.size'] = 6 * scaling
plt.rcParams['xtick.labelsize'] = 4 * scaling
plt.rcParams['ytick.labelsize'] = 4 * scaling
plt.rcParams['axes.labelsize'] = 6 * scaling
plt.rcParams['legend.fontsize'] = 5 * scaling
plt.rcParams['font.size'] = 7 * scaling
plt.rcParams['xtick.labelsize'] = 5 * scaling
plt.rcParams['ytick.labelsize'] = 5 * scaling
plt.rcParams['xtick.major.pad'] = -0.0 * scaling
plt.rcParams['ytick.major.pad'] = -0.0 * scaling
plt.rcParams['lines.linewidth'] = 0.65 * scaling
Expand Down Expand Up @@ -209,7 +220,8 @@
y_line = coefficients[0] * Xh + coefficients[1]

# Plot the linear line
ax.plot(Xh, y_line, label='lin', color='black', linestyle='--', linewidth=plt.rcParams['lines.linewidth']*0.8)
ax.plot(Xh, y_line, color='black', linestyle='--', linewidth=plt.rcParams['lines.linewidth']*0.8)
# ax.plot(Xh, y_line, label='lin', color='black', linestyle='--', linewidth=plt.rcParams['lines.linewidth']*0.8)

# Set legend & axis labels
ax.set_ylim(1e6, 4e6)
Expand Down Expand Up @@ -435,7 +447,7 @@ def interpolate_group(group):
ax.set_xlim(0, 100)
ax.set_xlabel("matched (%)")
ax.set_ylabel("elapsed time (s)")
ax.legend(handles=None, ncol=1, loc='upper left', fancybox=True, shadow=True)
ax.legend(handles=None, ncol=4, loc='upper left', fancybox=True, shadow=True, bbox_to_anchor=(1.0, 1.0))

# Ablation plots
filters = {
Expand Down Expand Up @@ -494,6 +506,7 @@ def interpolate_group(group):
}
plots_ablation_efficiency = {"combination_mode": None, "sort_mode": None, "backtrack": None}
plots_ablation_time = {"combination_mode": None, "sort_mode": None, "backtrack": None}
ncol = {"combination_mode": 1, "sort_mode": 1, "backtrack": 1}
for ablation_type in tqdm.tqdm(plots_ablation_efficiency.keys(), desc="Plotting ablation"):
d = df[run_mask[ablation_type]].copy()
# Only select data for this topology type and remove "topology_type" column from the dataframe
Expand All @@ -511,7 +524,44 @@ def interpolate_group(group):
ax.set_ylim(0, 100)
ax.set_xlabel('topology')
ax.set_ylabel("efficiency (%)")
ax.legend(handles=None, ncol=1, loc='best', fancybox=True, shadow=True)
if ablation_type == "backtrack":
from matplotlib.colors import LinearSegmentedColormap

# Create your custom colormap
cmap_colors = [(i, fcolor[i]) for i in [0, 5, 10, 15, 20]]
norm_vals = np.linspace(0, 20, len(cmap_colors))
colors = [(norm / 20, color) for norm, color in cmap_colors]
cmap = LinearSegmentedColormap.from_list('backtrack_cmap', colors)

_fig, _ax = plt.subplots(nrows=1, ncols=1, figsize=thirdwidth_figsize)
# cbar_ax = _fig.add_axes([0.2, 0.85, thirdwidth_figsize[0], thirdwidth_figsize[1]])
cbar_ax = fig.add_axes([0.16, 0.775, 0.67, 0.65])

# Hide grid lines
cbar_ax.grid(False)

# Hide axes ticks
cbar_ax.set_xticks([])
cbar_ax.set_yticks([])

# Hide axes
cbar_ax.set_frame_on(False)
cbar_ax.xaxis.set_visible(False)
cbar_ax.yaxis.set_visible(False)

# Generate some example heatmap data
data = np.random.randint(0, 21, (10, 10))
cax = _ax.matshow(data, cmap=cmap)

# Remove existing colorbar if any
if len(_fig.axes) > 1:
_fig.delaxes(_fig.axes[-1])

# Backtrack
cbar = plt.colorbar(cax, ax=cbar_ax, ticks=[0, 5, 10, 15, 20], orientation='horizontal', ticklocation='top')
ax.get_legend().remove()
else:
ax.legend(handles=None, ncol=ncol[ablation_type], loc='best', fancybox=True, shadow=True)

# Plot
fig, ax = plt.subplots(nrows=1, ncols=1, figsize=thirdwidth_figsize)
Expand All @@ -524,7 +574,45 @@ def interpolate_group(group):
ax.set_yscale("log")
ax.set_xlabel("topology")
ax.set_ylabel("elapsed time (s)")
ax.legend(handles=None, ncol=1, loc='best', fancybox=True, shadow=True)
if ablation_type == "backtrack":
from matplotlib.colors import LinearSegmentedColormap

# Create your custom colormap
cmap_colors = [(i, fcolor[i]) for i in [0, 5, 10, 15, 20]]
norm_vals = np.linspace(0, 20, len(cmap_colors))
colors = [(norm / 20, color) for norm, color in cmap_colors]
cmap = LinearSegmentedColormap.from_list('backtrack_cmap', colors)

_fig, _ax = plt.subplots(nrows=1, ncols=1, figsize=thirdwidth_figsize)
# cbar_ax = _fig.add_axes([0.2, 0.85, thirdwidth_figsize[0], thirdwidth_figsize[1]])
cbar_ax = fig.add_axes([0.16, 0.775, 0.67, 0.65])

# Hide grid lines
cbar_ax.grid(False)

# Hide axes ticks
cbar_ax.set_xticks([])
cbar_ax.set_yticks([])

# Hide axes
cbar_ax.set_frame_on(False)
cbar_ax.xaxis.set_visible(False)
cbar_ax.yaxis.set_visible(False)

# Generate some example heatmap data
data = np.random.randint(0, 21, (10, 10))
cax = _ax.matshow(data, cmap=cmap)

# Remove existing colorbar if any
if len(_fig.axes) > 1:
_fig.delaxes(_fig.axes[-1])

# Backtrack
cbar = plt.colorbar(cax, ax=cbar_ax, ticks=[0, 5, 10, 15, 20], orientation='horizontal', ticklocation='top')
ax.set_ylim([0, 140])
ax.get_legend().remove()
else:
ax.legend(handles=None, ncol=ncol[ablation_type], loc='best', fancybox=True, shadow=True)

# Performance plots
filters = {
Expand Down Expand Up @@ -568,7 +656,7 @@ def interpolate_group(group):
ax.set_ylim(0, 100)
ax.set_xlabel('nodes')
ax.set_ylabel("efficiency (%)")
ax.legend(handles=None, ncol=1, loc='upper right', fancybox=True, shadow=True)
ax.legend(handles=None, ncol=3, loc='upper right', fancybox=True, shadow=True, bbox_to_anchor=(2, 2))

# Plot performance_sigma
plots_perf_sigma = {"bidirectional-ring": None, "unidirectional-ring": None, "unirandom-ring": None}
Expand All @@ -590,7 +678,7 @@ def interpolate_group(group):
ax.set_ylim(0, 100)
ax.set_xlabel('sigma')
ax.set_ylabel("efficiency (%)")
ax.legend(handles=None, ncol=1, loc='upper right', fancybox=True, shadow=True)
ax.legend(handles=None, ncol=3, loc='upper right', fancybox=True, shadow=True, bbox_to_anchor=(2, 2))

# Save
PAPER_DIR = "/home/r2ci/Documents/project/MCS/MCS_RA-L/figures/python"
Expand All @@ -607,22 +695,52 @@ def interpolate_group(group):
if MUST_BREAK:
break
for topology_type, (fig, ax) in plots_transient.items():
fig.savefig(f"{PAPER_DIR}/transient_time_{topology_type}_legend{VERSION_ID}.pdf", bbox_inches=export_legend(fig, ax.get_legend()))
ax.get_legend().remove()
fig.savefig(f"{PAPER_DIR}/transient_time_{topology_type}{VERSION_ID}.pdf", bbox_inches='tight')
if MUST_BREAK:
break
for ablation_type, (fig, ax) in plots_ablation_efficiency.items():
fig.savefig(f"{PAPER_DIR}/ablation_efficiency_{ablation_type}{VERSION_ID}.pdf", bbox_inches='tight')
from matplotlib.transforms import Bbox, TransformedBbox
fig_width, fig_height = fig.get_size_inches()
width, height = thirdwidth_figsize # The dimensions you want for the lower-right corner in inches
# bbox = Bbox.from_bounds(0, 0, width, height)
bbox = ax.get_tightbbox(fig.canvas.get_renderer())
# Convert the bounding box to inches
bbox_inches = TransformedBbox(bbox, fig.dpi_scale_trans.inverted())
# bbox_inches = BBo
fig.savefig(
f"{PAPER_DIR}/ablation_efficiency_{ablation_type}{VERSION_ID}.pdf",
bbox_inches=bbox_inches
)
# fig.savefig(f"{PAPER_DIR}/ablation_efficiency_{ablation_type}{VERSION_ID}.pdf", bbox_inches='tight')
if MUST_BREAK:
break
for ablation_type, (fig, ax) in plots_ablation_time.items():
from matplotlib.transforms import Bbox, TransformedBbox
fig_width, fig_height = fig.get_size_inches()
width, height = thirdwidth_figsize # The dimensions you want for the lower-right corner in inches
# bbox = Bbox.from_bounds(0, 0, width, height)
bbox = ax.get_tightbbox(fig.canvas.get_renderer())
# Convert the bounding box to inches
bbox_inches = TransformedBbox(bbox, fig.dpi_scale_trans.inverted())
# bbox_inches = BBo
fig.savefig(
f"{PAPER_DIR}/ablation_time_{ablation_type}{VERSION_ID}.pdf",
bbox_inches=bbox_inches
)
fig.savefig(f"{PAPER_DIR}/ablation_time_{ablation_type}{VERSION_ID}.pdf", bbox_inches='tight')
if MUST_BREAK:
break
for topology_type, (fig, ax) in plots_perf_size.items():
fig.savefig(f"{PAPER_DIR}/performance_size_{topology_type}_legend{VERSION_ID}.pdf", bbox_inches=export_legend(fig, ax.get_legend()))
ax.get_legend().remove()
fig.savefig(f"{PAPER_DIR}/performance_size_{topology_type}{VERSION_ID}.pdf", bbox_inches='tight')
if MUST_BREAK:
break
for topology_type, (fig, ax) in plots_perf_sigma.items():
fig.savefig(f"{PAPER_DIR}/performance_sigma_{topology_type}_legend{VERSION_ID}.pdf", bbox_inches=export_legend(fig, ax.get_legend()))
ax.get_legend().remove()
fig.savefig(f"{PAPER_DIR}/performance_sigma_{topology_type}{VERSION_ID}.pdf", bbox_inches='tight')
if MUST_BREAK:
break

0 comments on commit ff4ec57

Please sign in to comment.