def select_interesting_navpoints(G, OD=None, N_per_sector=1, metric="centrality"):
"""
Select N_per_sector "interesting" navpoints per sector, according to
the metric given as input. The function selects the N_per_sector nodes
which have the higher metric within each sector.
Parameters
----------
G : hybrid network
OD : list of 2-tuples, optional
list of origin-destination nodes to use to compute the betweenness
centrality. If None is given, the betweenness is computed on all
possible pairs.
N_per_sector : int, optional
Number of interesting points to select per sector.
metric : string, optional
For now, only centrality is implemented.
Raises
------
Exception
If an unknown metric is given as input.
Returns
-------
n_best : dictionary
Keys are labels of sec-node and values are lists of labels of nav-nodes.
Notes
-----
Shall we compute the metric on the subnetwork of the sector?
"""
if metric=="centrality":
print "Computing betweenness centrality (between", len(OD), "pairs) ..."
bet = bet_OD(G.G_nav, OD=OD)
else:
raise Exception("Metric", metric, "is not implemented.")
# For each sector, sort the navpoints in increasing centrality and select the N_per_sector last
n_best = {sec:[G.node[sec]['navs'][idx] for idx in np.argsort([bet[nav] for nav in G.node[sec]['navs']])[-N_per_sector:]] for sec in G.nodes()}
return n_best
def select_interesting_navpoints_per_trajectory(trajs, G, OD=None, N_per_sec_per_traj=1, metric="centrality"):
"""
Equivalent of select_interesting_navpoints for trajectories.
Select N_per_sec_per_traj interesting napvoint per trajectory and per sector.
Parameters
----------
G : hybrid network
OD : list of 2-tuples, optional
list of origin-destination nodes to use to compute the betweenness
centrality. If None is given, the betweenness is computed on all
possible pairs.
N_per_sector : int, optional
Number of interesting points to select per sector.
metric : string, optional
For now, only centrality is implemented.
Raises
------
Exception
If an unknown metric is given as input.
Returns
-------
n_best : dictionary
Keys are labels of sec-node and values are lists of labels of nav-nodes.
"""
try:
assert hasattr(G, "G_nav")
except AssertionError:
raise Exception("Need an hybrid network (sectors+navpoints) in input.")
if metric=="centrality":
print "Computing betweenness centrality (between", len(OD), "pairs) ..."
bet = bet_OD(G.G_nav, OD=OD)
else:
raise Exception("Metric", metric, "is not implemented.")
n_best = {}
all_secs = set() # Only for information
for traj in trajs:
# Compute the sectors in trajectory
secs = set([G.G_nav.node[n]['sec'] for n in traj])
all_secs.union(secs)
#print "secs", secs
# For each sector, select the N_per_sec_per_traj best navpoints.
for sec in secs:
navs = [nav for nav in traj if G.G_nav.node[nav]['sec']==sec]
# print "navs:", navs
# print "bets:", [bet[nav] for nav in navs]
# print "idx:", np.argsort([bet[nav] for nav in navs])
# print "best:", [navs[idx] for idx in np.argsort([bet[nav] for nav in navs])[-N_per_sec_per_traj:]]
#for n in np.argsort([bet[nav] for nav in traj if G.G_nav.node[nav]['sec']==sec])[-N_per_sec_per_traj:]
n_best[sec] = n_best.get(sec, []) + [navs[idx] for idx in np.argsort([bet[nav] for nav in navs])[-N_per_sec_per_traj:]]
#print "n_best:", n_best
#print
# Remove redundant navpoints
n_best = {sec:list(set(navs)) for sec, navs in n_best.items()}
print "In average, I selected", np.mean([len(navs) for navs in n_best.values()]), "navpoint per sector."
return n_best
