def test_plot2d_graph():
df = pd.DataFrame([[1, 1], [6, 5], [6, 6], [0, 0], [1, 2], [2, 1], [5, 5],
[7, 6]])
pregraph = knn_graph(df, 3)
graph = pre_part_graph(pregraph, 10, df)
plot2d_graph(graph)
def test_pre_part_graph():
df = pd.DataFrame([[1, 1], [6, 5], [6, 6], [0, 0], [1, 2], [5, 5], [7, 2]])
k = 3
pregraph = knn_graph(df, k, verbose=False)
_ = pre_part_graph(pregraph, 2, df, verbose=False)
print(df["cluster"].values)
assert (df["cluster"].values == np.array([0, 1, 1, 0, 0, 1, 1])).all()
def test_rebuild_labels():
df = pd.DataFrame([[1, 1], [6, 5], [6, 6], [0, 0], [1, 2], [5, 5], [7, 2]])
pregraph = knn_graph(df, 3, verbose=False)
_ = pre_part_graph(pregraph, 10, df, verbose=False)
df_bis = rebuild_labels(df)
assert sorted(list(df_bis["cluster"])) == [1, 1, 1, 1, 2, 2, 2]
def test_len_edges():
df = pd.DataFrame([[1, 1], [6, 5], [6, 6], [0, 0], [1, 2], [5, 5], [7, 2]])
pregraph = knn_graph(df, 3, verbose=False)
graph = pre_part_graph(pregraph, 10, df, verbose=False)
condition0 = len_edges(graph, [0, 3, 4]) == 3
condition1 = len_edges(graph, [1, 2, 5, 6]) == 6
assert condition0 & condition1
def test_connecting_edges():
df = pd.DataFrame([[1, 1], [6, 5], [6, 6], [0, 0], [1, 2], [5, 5], [7, 2]])
pregraph = knn_graph(df, 3, verbose=False)
graph = pre_part_graph(pregraph, 10, df, verbose=False)
assert connecting_edges(([0, 3, 4], [1, 2, 5, 6]), graph) == [
(0, 5),
(3, 5),
(4, 5),
]
def test_get_cluster():
df = pd.DataFrame([[1, 1], [6, 5], [6, 6], [0, 0], [1, 2], [5, 5], [7, 2]])
k = 3
pregraph = knn_graph(df, 3, verbose=False)
graph = pre_part_graph(pregraph, k, df, verbose=False)
condition0 = get_cluster(graph, [0]) == [0, 3, 4]
condition1 = get_cluster(graph, [k - 1]) == [1, 2, 5, 6]
assert condition0 & condition1
def test_conn_comp():
df = pd.DataFrame([[1, 1], [6, 5], [6, 6], [0, 0], [1, 2], [2, 1], [5, 5],
[7, 6]])
pregraph = knn_graph(df, 3)
graph = pre_part_graph(pregraph, 10, df)
res = conn_comp(graph)
condition0 = sorted(res[0]) == [0, 3, 4, 5]
condition1 = sorted(res[1]) == [1, 2, 6, 7]
assert condition0 & condition1
def test_plot2d_data():
df = pd.DataFrame(make_blobs(60, random_state=42)[0])
knn = 6
pregraph = knn_graph(df, knn, verbose=False)
graph = pre_part_graph(pregraph, 10, df, verbose=False)
df, max_score, ci = merge_best(graph=graph,
df=df,
a=2,
k=3,
verbose=False,
verbose2=False)
plot2d_data(df)
def test_merge_best():
df = pd.DataFrame(make_blobs(60, random_state=42)[0])
knn = 6
pregraph = knn_graph(df, knn, verbose=False)
graph = pre_part_graph(pregraph, 10, df, verbose=False)
df, max_score, ci = merge_best(graph=graph,
df=df,
a=2,
k=3,
verbose=False,
verbose2=False)
condition0 = round(max_score) == 1
condition1 = ci == 5
assert condition0 & condition1
def test_knn_graph():
df = pd.DataFrame([[1, 1], [6, 5], [6, 6], [0, 0], [1, 2]])
k = 2
graph = knn_graph(df, k, verbose=False)
condition0 = list(graph.edges) == [
(0, 4),
(0, 3),
(1, 2),
(1, 4),
(2, 4),
(3, 4),
]
condition1 = list(graph.nodes) == [0, 1, 2, 3, 4]
assert condition0 & condition1
def cluster_gui(self, df, k, knn=10, m=30, alpha=2.0, save_plots=None):
if k is None:
k = 1
self.log.appendPlainText("Building kNN graph (k={})\n...".format(knn))
self.log.appendPlainText("")
graph = knn_graph(df, knn, False)
self.plot2d_graph_gui(
graph=graph,
canvas=self.canvas_up,
ax=self.ax1,
save_plots=save_plots,
ind_fig=self.ind_fig,
print_clust=False,
)
graph = self.pre_part_graph_gui(graph=graph,
canvas=self.canvas_up,
ax=self.ax1,
k=m,
df=df,
plotting=True)
# to account for cases where initial_clust is too big or k is already reached before the merging phase
cl_dict = {
list(graph.node)[i]: graph.node[i]["cluster"]
for i in range(len(graph))
}
m = len(Counter(cl_dict.values()))
self.log.appendPlainText("")
self.log.appendPlainText("actual init_clust: {}".format(m))
self.log.appendPlainText("")
dendr_height = OrderedDict({})
iterm = enumerate(range(m - k))
for i, _ in iterm:
df, ms, ci = self.merge_best_gui(graph,
df,
alpha,
k,
verbose=False,
verbose2=True)
if ms == 0:
break
dendr_height[m - (i + 1)] = ms
self.plot2d_data_gui(
df=df,
col_i=ci,
canvas=self.canvas_down,
ax=self.ax,
save_plots=save_plots,
ind_fig=self.ind_fig,
)
self.ind_fig += 1
res = rebuild_labels(df)
return res, dendr_height