def do_analysis(data, lens, name_prefix, nc, po, metric='euclidean'):
graph = mapper.map(lens,
data.values,
clusterer=lk.LinkageGap(verbose=0, metric=metric),
cover=km.Cover(n_cubes=nc, perc_overlap=po))
name = "{}_n{}_o{}".format(name_prefix, nc, po)
mapper.visualize(
graph,
# color_values=target.values,
color_values=lens,
color_function_name=name_prefix,
path_html=name + "_nm_diabetes_RM.html",
title=name + "nm_diabetes_RM")
nx_graph_simple = expo.kmapper_to_nxmapper(graph)
expo.cytoscapejson_dump(nx_graph_simple,
name + "_nm_diabetes_RM_simple.cyjs")
extra_data = {x: list(data.loc[:, x]) for x in data.columns}
extra_transforms = {
x: np.mean
for x in extra_data if x != "Clinical_classification"
}
nx_graph = expo.kmapper_to_nxmapper(graph,
node_extra_data=extra_data,
node_transforms=extra_transforms)
expo.cytoscapejson_dump(nx_graph, name + "_nm_diabetes_RM.cyjs")
def do_analysis(data, lens, name_prefix, nc, po):
name = "{}_n{}_o{}".format(name_prefix, nc, po)
graph = mapper.map(lens,
data,
clusterer=lk.LinkageGap(verbose=0),
cover=km.Cover(n_cubes=nc, perc_overlap=po))
mapper.visualize(graph,
color_function=lens,
path_html=name + "_cat.html",
title=name + "_cat")
def do_analysis(data, lens, cf, name_prefix, nc, po):
name = "{:s}_n{:d}_o{:.2f}".format(name_prefix, nc, po)
graph = mapper.map(lens,
data,
clusterer=lk.LinkageGap(verbose=0, metric='euclidean'),
cover=km.Cover(n_cubes=nc, perc_overlap=po))
mapper.visualize(graph,
color_function=cf,
path_html=name + "_travel.html",
title=name + "_travel")
def do_analysis(data, lens, name_prefix, nc, po,metric='euclidean'):
graph = mapper.map(lens,
data.values,
clusterer = lk.LinkageGap(verbose=0,metric=metric),
cover=km.Cover(n_cubes=nc, perc_overlap=po))
name = "{}_n{}_o{}".format(name_prefix,nc, po)
mapper.visualize(graph,
# color_function=target.values,
color_function=lens,
path_html=name + "_nm_diabetes_RM.html",
title=name + "nm_diabetes_RM")
def do_analysis(lens, name_prefix, nc, po):
graph = mapper.map(
lens,
data,
clusterer=lk.LinkageGap(verbose=0),
# clusterer=sklearn.cluster.DBSCAN(1, min_samples=0),
cover=km.Cover(n_cubes=nc, perc_overlap=po))
name = "{}_n{}_o{}".format(name_prefix, nc, po)
mapper.visualize(graph,
color_function=target,
path_html=name + "_iris.html",
title=name + "_iris")
def do_analysis(data, dists, lens, name_prefix, nc, po):
name = "{}_n{}_o{}".format(name_prefix, nc, po)
graph = mapper.map(lens,
dists,
clusterer=lk.LinkageGap(verbose=0,
metric="precomputed"),
precomputed=True,
cover=km.Cover(n_cubes=nc, perc_overlap=po))
mapper.visualize(graph,
color_values=lens,
color_function_name=name_prefix,
path_html=name + "_cat.html",
title=name + "_cat")
def do_analysis(data, lens, cf, name_prefix, nc, po):
name = "{:s}_n{:d}_o{:.2f}".format(name_prefix, nc, po)
graph = mapper.map(lens,
data.values,
clusterer=lk.LinkageGap(verbose=0, metric='euclidean'),
cover=km.Cover(n_cubes=nc, perc_overlap=po))
mapper.visualize(graph,
color_function=cf,
path_html=name + "_travel.html",
title=name + "_travel")
extra_data = {x: list(data.loc[:, x]) for x in data.columns}
extra_transforms = {x: np.mean for x in extra_data}
nx_graph = expo.kmapper_to_nxmapper(graph,
node_extra_data=extra_data,
node_transforms=extra_transforms)
expo.cytoscapejson_dump(nx_graph, name + "_travel.cyjs")
import kmapper as km
import mapperutils.linkage_gap as lk
import mapperutils.visualization as viz
mapper = km.KeplerMapper(verbose=2)
hand = trimesh.load_mesh("../0_data/hand/hand_simplified3k5.stl")
data = np.array(hand.vertices)
lens = data[:, 1:2]
plot = True
if plot:
viz.scatter3d(data, lens, colorsMap='viridis', show=False)
viz.plt.gca().view_init(elev=90, azim=0)
viz.axisEqual3D(viz.plt.gca())
viz.plt.show()
n = 7
p = 0.2
graph = mapper.map(lens,
data,
clusterer=lk.LinkageGap(verbose=0),
cover=km.Cover(n_cubes=n, perc_overlap=p))
name = "n{}_p{}".format(n, p)
mapper.visualize(graph,
color_function=lens,
path_html="hand_only_" + name + ".html",
title="hand, " + name)
def test_heuristics_precomputed():
dists = spd.squareform(spd.pdist(X))
fg = lk.LinkageGap(heuristic='firstgap', metric='precomputed').fit(dists)
assert len(np.unique(fg.labels_)) == 3
def test_heuristics():
fg = lk.LinkageGap(heuristic='firstgap').fit(X)
assert len(np.unique(fg.labels_)) == 3
with pytest.raises(Exception):
fg = lk.LinkageGap(heuristic='foobar')