Uni/boundary-aware-centrality
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

WIP diff centrality scores

Browse Source 
Check whether model correction is reliable in predicting the
"expected" outcome.
This commit is contained in:
Yves Biener
2026-03-21 21:16:18 +01:00
parent b323c724c9
commit ead3d70c35
1 changed files with 65 additions and 58 deletions
Download Patch File Download Diff File Expand all files Collapse all files

63
diff_comparison.py
View File

@@ -79,7 +79,7 @@ def spatial_graph(adata):
return g, weight return g, weight
def plot_graph_diff(G, c, fig, ax, name): def plot_graph_diff(G, c, fig, ax, name, cmap=plt.cm.plasma):
pos = G.vp["pos"] pos = G.vp["pos"]
x = [] x = []
y = [] y = []
@@ -89,7 +89,7 @@ def plot_graph_diff(G, c, fig, ax, name):
x.append(ver[0]) x.append(ver[0])
y.append(ver[1]) y.append(ver[1])
sc = ax.scatter(x, y, s=1, cmap=plt.cm.plasma, c=c) # map closeness values as color mapping on the verticies sc = ax.scatter(x, y, s=1, cmap=cmap, c=c) # map closeness values as color mapping on the verticies
ax.set_title(name) ax.set_title(name)
fig.colorbar(sc, ax=ax) fig.colorbar(sc, ax=ax)
@@ -200,7 +200,7 @@ row1, row2 = fig.subplots(2, 2)
ax1, ax2 = row1 ax1, ax2 = row1
# TODO select corresponding centrality measure method # TODO select corresponding centrality measure method
vp_closeness = apply(g, seed, weight, convex_hull, ax1, closeness, "Closeness") vp_closeness = apply(g, seed, weight, convex_hull, ax1, closeness, "Closeness")
# vp_betweenness = apply(g, seed, weight, convex_hull, ax2, betweenness, "Betweeness") vp_betweenness = apply(g, seed, weight, convex_hull, ax2, betweenness, "Betweeness")
# calculate convex hull # calculate convex hull
convex_hull = centrality.convex_hull(g_sub) convex_hull = centrality.convex_hull(g_sub)
@@ -214,20 +214,19 @@ fig_graph.savefig("Diff_subgraph.svg", format='svg')
ax1, ax2 = row2 ax1, ax2 = row2
vp_closeness_corrected = apply_corrected(g_sub, seed, weight_sub, convex_hull, ax1, closeness, "Closeness") vp_closeness_corrected = apply_corrected(g_sub, seed, weight_sub, convex_hull, ax1, closeness, "Closeness")
# vp_betweeness_corrected = apply_corrected(g_sub, seed, weight_sub, convex_hull, ax2, betweenness, "Betweeness") vp_betweeness_corrected = apply_corrected(g_sub, seed, weight_sub, convex_hull, ax2, betweenness, "Betweeness")
fig.savefig(f"Diff_scores.svg", format='svg') fig.savefig(f"Diff_scores.svg", format='svg')
# TODO how can I match the two vp's such that I can actually create a diff? for type in ['closeness', 'betweenness']:
# print(type)
print(f"Closeness: {vp_closeness}")
print(f"Closeness corrected: {vp_closeness_corrected}")
sub_keys = iter(g_sub.vertices()) sub_keys = iter(g_sub.vertices())
keys = iter(g.vertices()) keys = iter(g.vertices())
scores = [] scores = []
sub_scores = [] sub_scores = []
diff_scores = []
for sub_key in sub_keys: for sub_key in sub_keys:
key = next(keys) key = next(keys)
@@ -240,49 +239,57 @@ for sub_key in sub_keys:
# sub_position = g_sub.vp["pos"][sub_key] # sub_position = g_sub.vp["pos"][sub_key]
# print(f"position: {position} | sub_position: {sub_position}") # print(f"position: {position} | sub_position: {sub_position}")
value = 0.0
sub_value = 0.0
if type == 'closeness':
value = vp_closeness[key] value = vp_closeness[key]
sub_value = vp_closeness_corrected[sub_key] sub_value = vp_closeness_corrected[sub_key]
else:
value = vp_betweenness[key]
sub_value = vp_betweeness_corrected[sub_key]
scores.append(value) scores.append(value)
sub_scores.append(sub_value) sub_scores.append(sub_value)
# print(f"value: {value} | sub_value: {sub_value}") diff_scores.append(value - sub_value)
# TODO what do I want to know?
# - median score comparison?
# - max delta's between scores
# - improvement compared to with and without correction?
# TODO can I create the scatter graph with the points with their corresponding values?
median_score = np.median(scores) median_score = np.median(scores)
median_sub_score = np.median(sub_scores) median_sub_score = np.median(sub_scores)
print(f"median score: {median_score}") print(f"\tmedian score: {median_score}")
print(f"median sub_score: {median_sub_score}") print(f"\tmedian sub_score: {median_sub_score}")
print(f"median delta: {(median_score - median_sub_score)}") print(f"\tmedian delta: {(median_score - median_sub_score)}")
print("") print("")
max_value_score = np.max(scores) max_value_score = np.max(scores)
max_value_sub_score = np.max(sub_scores) max_value_sub_score = np.max(sub_scores)
print(f"max value score: {max_value_score}") print(f"\tmax value score: {max_value_score}")
print(f"max value sub_score: {max_value_sub_score}") print(f"\tmax value sub_score: {max_value_sub_score}")
print(f"max value delta: {(max_value_score - max_value_sub_score)}") print(f"\tmax value delta: {(max_value_score - max_value_sub_score)}")
print("") print("")
min_value_score = np.min(scores) min_value_score = np.min(scores)
min_value_sub_score = np.min(sub_scores) min_value_sub_score = np.min(sub_scores)
print(f"min value score: {min_value_score}") print(f"\tmin value score: {min_value_score}")
print(f"min value sub_score: {min_value_sub_score}") print(f"\tmin value sub_score: {min_value_sub_score}")
print(f"min value delta: {(min_value_score - min_value_sub_score)}") print(f"\tmin value delta: {(min_value_score - min_value_sub_score)}")
print("")
fig = plt.figure(figsize=(35, 10)) fig = plt.figure(figsize=(35, 10))
plot_graph_ax, plot_sub_graph_ax, plot_sub_graph_before_ax = fig.subplots(1, 3) plot_graph_ax, plot_sub_graph_ax, plot_sub_graph_before_ax = fig.subplots(1, 3)
plot_graph_diff(g, scores, fig, plot_graph_ax, "Original Graph (region of sub graph)") plot_graph_diff(g, scores, fig, plot_graph_ax, "Original Graph (region of sub graph)")
plot_graph_diff(g, sub_scores, fig, plot_sub_graph_ax, "Sub Graph (extracted region of original graph) with correction") plot_graph_diff(g, diff_scores, fig, plot_sub_graph_ax, "Differences after correction of sub graph compared to original graph", plt.cm.seismic)
vp = None
ep = None
if type == 'closeness':
vp = closeness(g_sub, weight=weight_sub) vp = closeness(g_sub, weight=weight_sub)
vp.a = np.nan_to_num(vp.a) # correct floating point values vp.a = np.nan_to_num(vp.a) # correct floating point values
else:
vp, ep = betweenness(g_sub, weight=weight_sub)
vp.a = np.nan_to_num(vp.a) # correct floating point values
# normalization # normalization
min_val, max_val = vp.a.min(), vp.a.max() # min_val, max_val = vp.a.min(), vp.a.max()
vp.a = (vp.a - min_val) / (max_val - min_val) # vp.a = (vp.a - min_val) / (max_val - min_val)
plot_graph_diff(g, vp.a, fig, plot_sub_graph_before_ax, "Sub Graph (extracted region of original graph) without correction") plot_graph_diff(g, vp.a, fig, plot_sub_graph_before_ax, "Sub Graph (extracted region of original graph) without correction")
fig.savefig(f"Diff_graph_scatter.svg", format='svg') fig.savefig(f"Diff_graph_scatter_{type}.svg", format='svg')