diff --git a/assignment-1/submission/19307130211/README.md b/assignment-1/submission/19307130211/README.md
new file mode 100755
index 0000000000000000000000000000000000000000..9d46bef52b53b558ad6ec716ee26dce570e77a87
--- /dev/null
+++ b/assignment-1/submission/19307130211/README.md
@@ -0,0 +1,213 @@
+
+
+[TOC]
+
+我只使用了numpy和matplotlib两个库实现了KNN的，所以代码应该可以完成在额外的任务。
+
+#### 一、模型的实现
+
+对于单个测试数据的knn分类网络的实现：
+
+```python
+inX=test_data[i]
+#计算欧氏距离
+diffMat=np.tile(inX,(dataSetSize,1))-tra_data
+sqDifMat=diffMat**2
+sqDistances=sqDifMat.sum(axis=1)
+distances=sqDistances**0.5
+#对计算得到的距离对应的序号进行排序
+sortedDistIndicies=distances.argsort()
+#统计最近的k个点中各个类别的数量
+classCount={}
+for j in range(k):
+	Label = tra_label[sortedDistIndicies[j]]
+	classCount[Label] = classCount.get(Label,0) + 1
+#得到对于类别的预测
+max_count=0
+for key,value in classCount.items():
+if value >max_count :
+	max_count = value
+	pred_label=key
+```
+
+对与KNN fit() 部分的实现：
+
+将数据集按照8:2的比例进行划分后，对K进行拟合，多次重复在训练集上进行单次的knn分类预测，选出正确率最高的K值。
+
+~~~python
+def fit(self, train_data, train_label):
+        self.train_data=train_data
+        self.train_label=train_label
+        N=train_data.shape[0]
+        cut=int(N*0.8)
+
+        tra_data, test_data = train_data[:cut,], train_data[cut:,]
+        tra_label, test_label = train_label[:cut,], train_label[cut:,]
+
+        dataSetSize=tra_data.shape[0]
+        test_number=test_data.shape[0]
+
+        best_k=0
+        max_score=0
+        if N<6 :
+            k_range=N
+        else :
+            k_range=20 
+
+        for k in range(2,k_range):
+            total_correct=0
+            for i in range(0,test_number):
+                inX=test_data[i]
+                #计算欧氏距离
+                diffMat=np.tile(inX,(dataSetSize,1))-tra_data
+                sqDifMat=diffMat**2
+                sqDistances=sqDifMat.sum(axis=1)
+                distances=sqDistances**0.5
+                #对计算得到的距离对应的序号进行排序
+                sortedDistIndicies=distances.argsort()
+                #统计最近的k个点中各个类别的数量
+                classCount={}
+                for j in range(k):
+                    Label = tra_label[sortedDistIndicies[j]]
+                    classCount[Label] = classCount.get(Label,0) + 1
+                #得到对于类别的预测
+                max_count=0
+                for key,value in classCount.items():
+                    if value >max_count :
+                        max_count = value
+                        pred_label=key
+                
+                if pred_label ==test_label[i]:
+                    total_correct=total_correct+1
+			#选择正确率最高的K值
+            score=total_correct*1.0/test_number
+            if score>max_score:
+                max_score=score
+                best_k=k
+        print("Best K: %d"%(best_k))
+        self.k=best_k
+~~~
+
+#### 二、实验数据的生成
+
+通过 np.random.multivariate_normal() 生成数据集
+
+~~~python
+	mean = (1, 1)
+    cov = np.array([[10, 0], [0, 10]])
+    x1 = np.random.multivariate_normal(mean, cov, (400,))
+
+    mean = (1, 12)
+    x2 = np.random.multivariate_normal(mean, cov, (400,))
+
+    mean = (12, 1)
+    x3 = np.random.multivariate_normal(mean, cov, (400,))
+
+    mean = (12, 12)
+    x4 = np.random.multivariate_normal(mean, cov, (400,))
+~~~
+
+生成的训练集：![Figure_1](img/Figure_1.png)
+
+生成的测试集：![Figure_2](img/Figure_2.png)
+
+训练结果：
+
+| K    | 准确率   |
+| ---- | :------- |
+| 10   | 0.884375 |
+
+#### 三、探究实验
+
+##### 1. 修改mean参数
+
+首先修改成如下形式，使得四类点分布的更开
+
+~~~python
+mean = (1, 1) mean = (1, 15) mean = (15, 1) mean = (15, 15)
+~~~
+
+得到的训练集：![Figure_3](img/Figure_3.png)
+
+得到的测试集：![Figure_4](img/Figure_4.png)
+
+训练结果：
+
+| K    | 准确率   |
+| ---- | -------- |
+| 11   | 0.978125 |
+
+可以看到不同的数据分布更开后，准确率大大提升。
+
+
+
+修改成以下的形式后，不同的数据更加集中：
+
+~~~python
+ean = (1, 1) mean = (1, 8) mean = (8, 1) mean = (8, 8)
+~~~
+
+生成的训练集：![Figure_5](img/Figure_5.png)
+
+生成的测试集：![Figure_6](img/Figure_6.png)
+
+训练结果：
+
+| K    | 准确率 |
+| ---- | ------ |
+| 18   | 0.7    |
+
+可以看到准确率大大降低。
+
+结论：KNN分类算法的准确性依赖于数据的分布，当不同种类的数据越集中（重合的部分越多时），效果越差；
+
+反之，重合部分越少，分类结果越好。
+
+##### 2.对cov参数的修改
+
+修改cov,使得方差更大，同类数据更离散:
+
+~~~python
+cov = np.array([[30, 0], [0, 30]])
+~~~
+
+生成的训练集：![Figure_7](img/Figure_7.png)
+
+生成的测试集：![Figure_8](img/Figure_8.png)
+
+测试结果：
+
+| K    | 准确率   |
+| ---- | -------- |
+| 18   | 0.690625 |
+
+可以看到数据更离散后，准确率下降
+
+
+
+再次修改cov，减少方差，使得同类数据更集中
+
+```python
+cov = np.array([[5, 0], [0, 5]])
+```
+
+生成的训练集：![Figure_9](img/Figure_9.png)
+
+生成的测试集：![Figure_10](img/Figure_10.png)
+
+训练结果：
+
+| K    | 准确率 |
+| ---- | ------ |
+| 5    | 0.9875 |
+
+可以看到同类数据更集中，准确率越高。
+
+##### 探讨与结论
+
+可以和实验一得到相同的结论，不同类数据间分布的重合区域越小，KNN分类方法的准确率越高；重合区域越大，准确率越低。
+
+
+
+同时在上述实验中观察到的另一有趣现象是准确率越低时，K的最佳取指偏大；准确率越高时，K的取值偏小。但是在修改参数cov来控制正确率，希望得到准确率和K的关系时，却发现最佳K的选取还是具有一定的随机性，没有找到一般规律。
+
diff --git a/assignment-1/submission/19307130211/img/Figure_1.png b/assignment-1/submission/19307130211/img/Figure_1.png
new file mode 100755
index 0000000000000000000000000000000000000000..e4a5d05eb4f6ccb904e8fd207fded8cf0b68b43c
Binary files /dev/null and b/assignment-1/submission/19307130211/img/Figure_1.png differ
diff --git a/assignment-1/submission/19307130211/img/Figure_10.png b/assignment-1/submission/19307130211/img/Figure_10.png
new file mode 100755
index 0000000000000000000000000000000000000000..c9002a2fdce20748784f39dfaac3a0e1ca956f06
Binary files /dev/null and b/assignment-1/submission/19307130211/img/Figure_10.png differ
diff --git a/assignment-1/submission/19307130211/img/Figure_2.png b/assignment-1/submission/19307130211/img/Figure_2.png
new file mode 100755
index 0000000000000000000000000000000000000000..c4cb6a263cd63039ad370eca82607f50666592b7
Binary files /dev/null and b/assignment-1/submission/19307130211/img/Figure_2.png differ
diff --git a/assignment-1/submission/19307130211/img/Figure_3.png b/assignment-1/submission/19307130211/img/Figure_3.png
new file mode 100755
index 0000000000000000000000000000000000000000..24fc664d1ea502c02c96a4deca581dc42fc465cd
Binary files /dev/null and b/assignment-1/submission/19307130211/img/Figure_3.png differ
diff --git a/assignment-1/submission/19307130211/img/Figure_4.png b/assignment-1/submission/19307130211/img/Figure_4.png
new file mode 100755
index 0000000000000000000000000000000000000000..1a3fa01d1aa274fb14dca65e961793de8f2174ca
Binary files /dev/null and b/assignment-1/submission/19307130211/img/Figure_4.png differ
diff --git a/assignment-1/submission/19307130211/img/Figure_5.png b/assignment-1/submission/19307130211/img/Figure_5.png
new file mode 100755
index 0000000000000000000000000000000000000000..4afbaeb755e3460b5f5e547d8e43e5221cd44c17
Binary files /dev/null and b/assignment-1/submission/19307130211/img/Figure_5.png differ
diff --git a/assignment-1/submission/19307130211/img/Figure_6.png b/assignment-1/submission/19307130211/img/Figure_6.png
new file mode 100755
index 0000000000000000000000000000000000000000..4dd7b5c0d0d15d0bfb3f47c5db12a5a378e17144
Binary files /dev/null and b/assignment-1/submission/19307130211/img/Figure_6.png differ
diff --git a/assignment-1/submission/19307130211/img/Figure_7.png b/assignment-1/submission/19307130211/img/Figure_7.png
new file mode 100755
index 0000000000000000000000000000000000000000..89613e8f29af003f9ad06af59112d5199da8d5a9
Binary files /dev/null and b/assignment-1/submission/19307130211/img/Figure_7.png differ
diff --git a/assignment-1/submission/19307130211/img/Figure_8.png b/assignment-1/submission/19307130211/img/Figure_8.png
new file mode 100755
index 0000000000000000000000000000000000000000..87a4ebbe6926fe0a26cc6e91c6c31e96b769736b
Binary files /dev/null and b/assignment-1/submission/19307130211/img/Figure_8.png differ
diff --git a/assignment-1/submission/19307130211/img/Figure_9.png b/assignment-1/submission/19307130211/img/Figure_9.png
new file mode 100755
index 0000000000000000000000000000000000000000..dff482725b27547590be277cdba748ba1bc1305f
Binary files /dev/null and b/assignment-1/submission/19307130211/img/Figure_9.png differ
diff --git a/assignment-1/submission/19307130211/source.py b/assignment-1/submission/19307130211/source.py
new file mode 100755
index 0000000000000000000000000000000000000000..43e2a2ec214289e40b5a5c7e99b49087695fc828
--- /dev/null
+++ b/assignment-1/submission/19307130211/source.py
@@ -0,0 +1,154 @@
+import numpy as np
+import matplotlib.pyplot as plt
+
+class KNN:
+
+    def __init__(self):
+        pass
+    
+    #def settled_knn(self,test_data,train_data,train_label) :
+
+    def fit(self, train_data, train_label):
+        self.train_data=train_data
+        self.train_label=train_label
+        N=train_data.shape[0]
+        cut=int(N*0.8)
+
+        tra_data, test_data = train_data[:cut,], train_data[cut:,]
+        tra_label, test_label = train_label[:cut,], train_label[cut:,]
+
+        dataSetSize=tra_data.shape[0]
+        test_number=test_data.shape[0]
+
+        best_k=0
+        max_score=0
+        if N<6 :
+            k_range=N
+        else :
+            k_range=20 
+
+        for k in range(2,k_range):
+            total_correct=0
+            for i in range(0,test_number):
+                inX=test_data[i]
+
+                diffMat=np.tile(inX,(dataSetSize,1))-tra_data
+
+                sqDifMat=diffMat**2
+                sqDistances=sqDifMat.sum(axis=1)
+                distances=sqDistances**0.5
+
+                sortedDistIndicies=distances.argsort()
+
+                classCount={}
+                for j in range(k):
+                    Label = tra_label[sortedDistIndicies[j]]
+                    classCount[Label] = classCount.get(Label,0) + 1
+                max_count=0
+                for key,value in classCount.items():
+                    if value >max_count :
+                        max_count = value
+                        pred_label=key
+                
+                if pred_label ==test_label[i]:
+                    total_correct=total_correct+1
+
+            score=total_correct*1.0/test_number
+            if score>max_score:
+                max_score=score
+                best_k=k
+        print("Best K: %d"%(best_k))
+        self.k=best_k
+
+    def predict(self, test_data):
+        dataSetSize=self.train_data.shape[0]
+        test_number=test_data.shape[0]
+        ans_label=np.array([])
+        for i in range(0,test_number):
+                inX=test_data[i]
+
+                diffMat=np.tile(inX,(dataSetSize,1))-self.train_data
+
+                sqDifMat=diffMat**2
+                sqDistances=sqDifMat.sum(axis=1)
+                distances=sqDistances**0.5
+
+                sortedDistIndicies=distances.argsort()
+                classCount={}
+                for j in range(self.k):
+                    voteIlabel = self.train_label[sortedDistIndicies[j]]
+                    classCount[voteIlabel] = classCount.get(voteIlabel,0) + 1
+
+                max_count=0
+                pred_label=self.train_label[0]
+                for key,value in classCount.items():
+                    if value >max_count :
+                        max_count = value
+                        pred_label=key
+
+                ans_label=np.append(ans_label,pred_label)
+        return ans_label
+        
+def generate():
+    mean = (1, 1)
+    cov = np.array([[10, 0], [0, 10]])
+    x1 = np.random.multivariate_normal(mean, cov, (400,))
+
+    mean = (1, 12)
+    cov = np.array([[10, 0], [0, 10]])
+    x2 = np.random.multivariate_normal(mean, cov, (400,))
+
+    mean = (12, 1)
+    cov = np.array([[10, 0], [0, 10]])
+    x3 = np.random.multivariate_normal(mean, cov, (400,))
+
+    mean = (12, 12)
+    cov = np.array([[10, 0], [0, 10]])
+    x4 = np.random.multivariate_normal(mean, cov, (400,))
+
+    X = np.concatenate([x1, x2, x3, x4])
+    Y = np.concatenate([
+        np.zeros((400,),dtype=int),
+        np.ones((400,),dtype=int),
+        np.ones((400,),dtype=int)*2,
+        np.ones((400,),dtype=int)*3
+    ])
+    shuffled_indices = np.random.permutation(1600)
+    data=X[shuffled_indices]
+    label=Y[shuffled_indices]
+    total=1600
+    cut=int(total*0.8)
+    train_data, test_data = data[:cut,], data[cut:,]
+    train_label, test_label = label[:cut,], label[cut:,]
+
+    np.save("data.npy",(
+         (train_data, train_label), (test_data, test_label)
+    ))
+
+def read():
+    (train_data, train_label), (test_data, test_label) = np.load("data.npy",allow_pickle=True)
+    return (train_data, train_label), (test_data, test_label)
+
+def display(data,label,name):
+    datas=[[],[],[],[]]
+    for i in range(len(data)):
+        
+        datas[label[i]].append(data[i])
+
+    for each in datas:
+        each=np.array(each)
+        plt.scatter(each[:,0],each[:,1])
+    #plt.savefig(f'imag/{name}')
+    plt.show()
+
+
+if __name__ == '__main__':
+    generate()
+    (train_data, train_label), (test_data, test_label) = read()
+    display(train_data, train_label, 'train')
+    display(test_data, test_label, 'test')
+
+    model = KNN()
+    model.fit(train_data, train_label)
+    res = model.predict(test_data)
+    print("acc =",np.mean(np.equal(res, test_label)))
\ No newline at end of file