From ef3fdca284b5900ba61e107d5bb6e7e80cdaad0e Mon Sep 17 00:00:00 2001
From: Cextel <2200221005@email.szu.edu.cn>
Date: Wed, 28 Feb 2024 23:28:49 +0800
Subject: [PATCH] =?UTF-8?q?=E4=BF=AE=E6=94=B9Seq=5Fto=5Fvec=E5=87=BD?=
 =?UTF-8?q?=E6=95=B0?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

最后的向量转换由for循环转换为np.mean方法，以提高批量计算时的速度。
---
 README.md | 8 +++-----
 1 file changed, 3 insertions(+), 5 deletions(-)

diff --git a/README.md b/README.md
index 9ad4ede..3e41d99 100644
--- a/README.md
+++ b/README.md
@@ -191,11 +191,9 @@ def Seq_to_vec(Sequence):
             seq_emd = embedding[seq_num][:seq_len - 1]
             features.append(seq_emd)
     features_normalize = np.zeros([len(features), len(features[0][0])], dtype=float)
-    for i in range(len(features)):
-        for k in range(len(features[0][0])):
-            for j in range(len(features[i])):
-                features_normalize[i][k] += features[i][j][k]
-            features_normalize[i][k] /= len(features[i])
+    features_array = np.array(features)
+    for n in range(len(features)):
+        features_normalize[n] = features_array[n].mean(axis=0, dtype=np.float64)
     return features_normalize