diff --git a/python/.keep b/python/.keep
new file mode 100644
index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391
diff --git a/python/python.md b/python/python.md
new file mode 100644
index 0000000000000000000000000000000000000000..17890621b2d100689a721bcaf8f79855e99db771
--- /dev/null
+++ b/python/python.md
@@ -0,0 +1,55 @@
+方法的默认参数被设置为可变对象，导致调用方法时参数没有初始化，被反复利用  
+
+```python
+def train_knnm(x, labels, erd, representatives=list()):
+    states = np.zeros(len(x))  # Mark all the instance as 'not covered'
+    distance_matrix = get_distance_matrix(x)
+    not_covered = get_not_covered(states)
+
+    # while there are examples to group
+    while len(not_covered) > 0:
+        max_neighbourhood = list()
+        global min_md
+        tuple_max_neighbourhood = None
+        for i in not_covered:
+            # get distance from i to all other tuples
+            distances = distance_matrix[i]
+
+            # sort distances
+            sorted_distances = [d for d in sorted(enumerate(distances), key=lambda a: a[1])]
+
+            # filter only those which has not been yet covered
+            sorted_distances = [d for d in sorted_distances if states[d[0]] == 0]
+
+            # compute neighbourhood
+            q = 0
+            neighbourhood = list()
+            errors = 0
+            while q < len(sorted_distances) and (labels[sorted_distances[q][0]] == labels[i] or errors < erd):
+                neighbourhood.append(sorted_distances[q][0])
+                if labels[sorted_distances[q][0]] != labels[i]:
+                    errors += 1
+                q += 1
+            if len(neighbourhood) > len(max_neighbourhood):#找当前最大的簇
+                max_neighbourhood = neighbourhood
+                tuple_max_neighbourhood = i
+        # add representative
+        # representatives format (rep(di), all_tuples in neighbourhood, class(di), Sim(di))
+        rep = (tuple_max_neighbourhood, x[tuple_max_neighbourhood])
+        num = max_neighbourhood
+        cls = labels[tuple_max_neighbourhood]
+        sim = distance_matrix[tuple_max_neighbourhood, max_neighbourhood[-1]]
+        lay = 0
+        representatives.append([rep, num, cls, sim, lay])#当前最大的簇形成新簇
+        if len(num)/sim<min_md :
+            min_md=len(num)/sim
+        # update states array
+        for i in max_neighbourhood:#更新覆盖状态
+            states[i] = 1
+        not_covered = get_not_covered(states)
+
+        # print(len(representatives), len(representatives[len(representatives)-1][1]))
+
+    return representatives
+
+```
\ No newline at end of file