fast !

function.py

@@ -39,7 +39,11 @@ name = "向兴宇"
 
 
 # 这里写自己的代码实现
-def classify0(inX, dataSet, labels, k=8, _cache={"VI": None, "dataSet": None}):
+import numba
+
+
+# @numba.jit(nopython=True, fastmath=True)
+def classify0(inX, dataSet, labels, k=8, _cache={"VI": None}):
     """
     KNN 分类算法。
 
@@ -61,16 +65,13 @@ def classify0(inX, dataSet, labels, k=8, _cache={"VI": None, "dataSet": None}):
     diff = dataSet - inX
 
     # 使用矩阵运算计算马氏距离
-    distances = np.sum(diff @ _cache["VI"] * diff, axis=1)
-
-    # 对距离排序，获取前 k 个最近的样本索引
-    sortedDistIndices = np.argsort(distances)
+    # distances = np.einsum("ij,jk,ik->i", diff, _cache["VI"], diff)
 
     # 统计前 k 个样本的标签
-    classCount = np.bincount(labels[sortedDistIndices[:k]])
+    # classCount = np.bincount(labels[np.argpartition(distances, k)[:k]])
 
-    # 返回出现次数最多的标签
-    return np.argmax(classCount)
+    # 究极堆叠，减少赋值操作
+    return np.argmax(np.bincount(labels[np.argpartition(np.einsum("ij,jk,ik->i", diff, _cache["VI"], diff), k)[:k]]))
 
 
 def load_data(file_path):