diff --git a/assignment-3/submission/17307130331/README.md b/assignment-3/submission/17307130331/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..346d3af489c96fe167b2c767a91caff1197e158a
--- /dev/null
+++ b/assignment-3/submission/17307130331/README.md
@@ -0,0 +1,62 @@
+# 实验报告
+
+**17307130331 陈疏桐**
+
+本次实验实现了Kmeans聚类和GaussianMixture聚类模型，并生成数据完成了基本实验。模型代码见`source.py`。
+
+## 基本实验
+
+### 数据集1
+
+随机产生二元高斯分布数据，数据集的详细情况如下：
+
+```
+CLUSTER-1:
+  mean = (0,0)
+  variance = ([[124,0],[0,225]])
+CLUSTER-2:
+  mean = (-35, 40)
+  variance = ([[32,0],[0,33]])
+CLUSTER-3:
+  mean = ((30, 50))
+  variance = ([[44, 0], [0, 20]])
+```
+
+
+KMeans和GaussianMixture聚类结果分别如下：
+![km1](img/p1.png)
+![gs1](img/p2.png)
+
+图中不同颜色代表数据点的不同标签；红色点是簇的中心。由图中可见，在三个簇交汇的边缘，KMeans以距离中心的远近判断数据点所属的簇，因此在簇的边缘将数据点分到三个不同的簇，GaussianMixture则有更好的聚类效果。
+
+### 数据集2 
+
+同样是2元高斯分布，数据集情况如下：
+```
+CLUSTER-1:
+    mean = (0,0)
+    variance = ([[124,0],[0,225]])
+CLUSTER-2:
+    mean = (-20, -25)
+    variance = [[132,0],[0,33]]
+CLUSTER-3:
+    mean = (30, 15)
+    variance = ([[73, 0], [0, 20]])
+CLUSTER-4:
+    mean = (11, -2)
+    variance = ([[34,0], [0, 138]])
+```
+
+KMeans和GaussianMixture聚类结果分别如下：
+![km1](img/p3.png)
+![gs1](img/p4.png)
+可见，GaussianMixture的聚类效果还是比KMeans要更好，聚类中心与数据集真实分布的均值更接近；另外，KMeans聚类的边界更加规则、清晰，GaussianMixture更适合聚类之间分布重叠程度更大的、呈椭圆状分布的数据。
+
+### 数据集3
+
+利用`scikit-learn`产生环状簇聚数据集。
+用KMeans 和 GaussianMixture进行聚类，结果分别如图：
+![km2](img/p7.png)
+![gs3](img/p8.png)
+
+对于这种非高斯分布、非球形或椭圆形的数据，KMeans和GaussianMixture聚类的效果差不多，且都不理想，可以用其他方式，如基于密度的聚类方法来聚类。
\ No newline at end of file
diff --git a/assignment-3/submission/17307130331/img/p1.png b/assignment-3/submission/17307130331/img/p1.png
new file mode 100644
index 0000000000000000000000000000000000000000..e3f8ddb046e5868f6b3b82a7f421a6fa9b1fa12d
Binary files /dev/null and b/assignment-3/submission/17307130331/img/p1.png differ
diff --git a/assignment-3/submission/17307130331/img/p2.png b/assignment-3/submission/17307130331/img/p2.png
new file mode 100644
index 0000000000000000000000000000000000000000..95e2926b9ef5be60d2de0e23721349c5dfd0a57d
Binary files /dev/null and b/assignment-3/submission/17307130331/img/p2.png differ
diff --git a/assignment-3/submission/17307130331/img/p3.png b/assignment-3/submission/17307130331/img/p3.png
new file mode 100644
index 0000000000000000000000000000000000000000..8178173f9b681041311f5c853f43e44cbdc4fd6d
Binary files /dev/null and b/assignment-3/submission/17307130331/img/p3.png differ
diff --git a/assignment-3/submission/17307130331/img/p4.png b/assignment-3/submission/17307130331/img/p4.png
new file mode 100644
index 0000000000000000000000000000000000000000..aac3b637e7c0d75304ee0f360134c91fc7b696e0
Binary files /dev/null and b/assignment-3/submission/17307130331/img/p4.png differ
diff --git a/assignment-3/submission/17307130331/img/p5.png b/assignment-3/submission/17307130331/img/p5.png
new file mode 100644
index 0000000000000000000000000000000000000000..cf11b621975a9f5165372ccb8bd33d4bad6c1b62
Binary files /dev/null and b/assignment-3/submission/17307130331/img/p5.png differ
diff --git a/assignment-3/submission/17307130331/img/p6.png b/assignment-3/submission/17307130331/img/p6.png
new file mode 100644
index 0000000000000000000000000000000000000000..d6a042d33ad9b3c2c6c542146eee165414e180a6
Binary files /dev/null and b/assignment-3/submission/17307130331/img/p6.png differ
diff --git a/assignment-3/submission/17307130331/img/p7.png b/assignment-3/submission/17307130331/img/p7.png
new file mode 100644
index 0000000000000000000000000000000000000000..28553042f8529afa6f8f0f3bfe5300c95e27aee5
Binary files /dev/null and b/assignment-3/submission/17307130331/img/p7.png differ
diff --git a/assignment-3/submission/17307130331/img/p8.png b/assignment-3/submission/17307130331/img/p8.png
new file mode 100644
index 0000000000000000000000000000000000000000..7c4f7aa08fb25490ef28f00aaa60a4d12d4fd9c3
Binary files /dev/null and b/assignment-3/submission/17307130331/img/p8.png differ
diff --git a/assignment-3/submission/17307130331/source.py b/assignment-3/submission/17307130331/source.py
new file mode 100644
index 0000000000000000000000000000000000000000..b3c6e3de22e84a46fbaa5e4b25f5b7cf23ac34ce
--- /dev/null
+++ b/assignment-3/submission/17307130331/source.py
@@ -0,0 +1,243 @@
+import numpy as np
+import random
+import matplotlib.pyplot as plt
+
+class KMeans:
+
+    def __init__(self, n_clusters, max_iters=None):
+        self.n_clusters = n_clusters
+        self.centers = None
+        self.n = None
+        self.max_iters = max_iters
+    
+    def initial_centers(self, train_data):
+        #idx = [i for i in range(train_data.shape[0])]
+        #rdx = random.sample(idx, self.n_clusters)
+        rng = np.random.default_rng()
+        centers = rng.choice(train_data, size=self.n_clusters, axis=0, replace=False) # 无放回抽样
+        #centers = train_data[rdx]
+        return centers
+    
+    
+    def dis_matrix(self, x, y, square=True):
+        
+        xx = np.einsum('ij,ij->i', x, x)[:, np.newaxis]  
+        yy = np.einsum('ij,ij->i', y, y)[np.newaxis, :]  
+        dot = np.dot(x, y.T)  
+        res = np.abs(xx + yy - 2 * dot)  # (x-y)^2=x^2+y^2-2xy
+        return res if square else np.sqrt(res + 1e-9)
+    
+    def avg_dis(self, dots):
+        dis_sum = np.einsum("ij->", self.dis_matrix(dots, dots, False))
+        return dis_sum/(2*dots.shape[0])
+    
+    def _iter_(self, data):
+        cluster_labels = np.argmin(self.dis_matrix(data, self.centers), axis=1) # 把每个数据点分配到距离最近的中心所属簇
+        clusters = [[] for _ in range(self.n_clusters)]    # 保存各个簇中的数据点，方便再计算簇中心
+        for i in range(data.shape[0]):
+            clusters[cluster_labels[i]].append(data[i])
+        for i in range(self.n_clusters):
+            clusters[i] = np.array(clusters[i])
+        return cluster_labels, clusters
+    
+    def fit(self, train_data):
+        self.n = train_data.shape[0]
+        self.centers = self.initial_centers(train_data)
+        cluster_labels, clusters = self._iter_(train_data)
+        
+        num_iters = 0
+        
+        while True:
+            num_iters += 1
+            self.centers = np.array([np.mean(c, axis=0) for c in clusters])
+            new_cluster_labels, clusters = self._iter_(train_data)
+            if (new_cluster_labels == cluster_labels).all():
+                break
+            if self.max_iters and num_iters == self.max_iters:
+                break
+            cluster_labels = new_cluster_labels
+        return sum([self.avg_dis(c) for c in clusters])         
+    
+    def predict(self, test_data):
+        assert self.centers is not None
+        return np.argmin(self.dis_matrix(test_data, self.centers), axis=1)
+
+class GaussianMixture:
+
+    def __init__(self, n_clusters, max_iters=50):
+        self.n_clusters = n_clusters # 聚类簇数
+        self.n = None # 训练样本容量
+        self.dim = None # 样本维度
+        self.cov = None # 协方差
+        self.mean = None # 高斯分布的均值
+        self.max_iters = max_iters
+        self.pi = [1/self.n_clusters for i in range(self.n_clusters)]
+        self.gamma = None
+        
+    def prob_normal(self, data, mean, cov):
+        tmp = (data-mean) @ np.linalg.inv(cov) @ (data-mean).T
+        tmp2 = (2 * np.pi) ** (-len(data)/2) * np.linalg.det(cov)**(-1/2)
+        return np.exp(tmp /(-2)) * tmp2
+        
+    
+    def dis_matrix(self, x, y, square=True):
+
+        xx = np.einsum('ij,ij->i', x, x)[:, np.newaxis]  
+        yy = np.einsum('ij,ij->i', y, y)[np.newaxis, :]  
+        dot = np.dot(x, y.T)  
+        res = np.abs(xx + yy - 2 * dot)  # (x-y)^2=x^2+y^2-2xy
+        return res if square else np.sqrt(res + 1e-9)
+    
+    def avg_dis(self, dots):
+        dis_sum = np.einsum("ij->", self.dis_matrix(dots, dots, False))
+        return dis_sum/(2*dots.shape[0])
+    def initial_centers(self, train_data):
+        #idx = [i for i in range(train_data.shape[0])]
+        #rdx = random.sample(idx, self.n_clusters)
+        rng = np.random.default_rng()
+        centers = rng.choice(train_data, size=self.n_clusters, axis=0, replace=False) # 无放回抽样
+        dis_centers = self.avg_dis(centers)
+        for i in range(10):
+            new_centers = rng.choice(train_data, size=self.n_clusters, axis=0, replace=False)
+            new_dis = self.avg_dis(new_centers)
+            if new_dis>dis_centers:
+                dis_centers = new_dis
+                centers = new_centers
+            
+        #centers = train_data[rdx]
+        return centers
+        
+    
+    def fit(self, train_data):
+        
+        self.n = train_data.shape[0]
+        self.dim = train_data.shape[1]
+        
+        if train_data.shape[0] < 5 * train_data.shape[1]:
+            train_data = np.concatenate([train_data, train_data+np.random.rand(self.n, self.dim),
+                                        train_data+np.random.rand(self.n, self.dim),
+                                        train_data+np.random.rand(self.n, self.dim),
+                                        train_data+np.random.rand(self.n, self.dim)], axis=0)
+        self.n = train_data.shape[0]
+        self.dim = train_data.shape[1]
+        
+        # 初始化均值和方差
+        self.mean = self.initial_centers(train_data)
+        self.cov = np.concatenate([np.eye(self.dim)[np.newaxis, :] for _ in range(self.n_clusters)], axis=0)
+        
+        num_iters = 0
+        
+        while True:
+            num_iters+=1
+            # E step
+            # 计算每一个数据点属于每一个分布的概率
+            self.gamma = np.zeros(shape = (self.n, self.n_clusters))
+            for i in range(self.n):
+                for j in range(self.n_clusters):
+                    self.gamma[i,j] = self.pi[j] * self.prob_normal(train_data[i], self.mean[j], self.cov[j])
+                    self.gamma[i,j] /= sum([self.pi[k] * self.prob_normal(train_data[i], self.mean[k], self.cov[k]) for k in range(self.n_clusters)])+1e-10
+            N = np.sum(self.gamma, axis=0)
+            
+            # M step
+            # 更新均值
+            prev_mean = self.mean
+            self.mean = np.zeros(shape = (self.n_clusters, self.dim))
+            for i in range(self.n_clusters):
+                self.mean[i] = np.average(train_data, axis=0, weights=self.gamma[:, i])
+                
+            # 更新方差
+            prev_cov = self.cov
+            for i in range(self.n_clusters):
+                tmp = np.array([np.dot((train_data - self.mean[i])[j].reshape(-1,1), (train_data-self.mean[i])[j].reshape(1, -1)) for j in range(self.n)]) 
+                self.cov[i] = np.average(tmp, axis=0, weights = self.gamma[:, i])
+                #self.cov[i] /= N[i]
+
+            # 更新pi
+            self.pi = [N[i]/self.n for i in range(self.n_clusters)]
+            
+            if self.max_iters and num_iters==self.max_iters:
+                break
+            if (prev_mean==self.mean).all() and (prev_cov == self.cov).all():
+                break
+        
+    
+    def predict(self, test_data):
+        assert self.mean is not None
+        assert self.cov is not None
+        probs = []
+        
+        for i in range(test_data.shape[0]):
+            probs.append([self.prob_normal(test_data[i], self.mean[k], self.cov[k]) for k in range(self.n_clusters)])
+        return np.argmax(probs, axis=1)
+
+class ClusteringAlgorithm:
+
+    def __init__(self):
+        pass
+    
+    def fit(self, train_data):
+        pass
+    
+    def predict(self, test_data):
+        pass
+    
+def shuffle(*datas):
+    data = np.concatenate(datas)
+    label = np.concatenate([
+        np.ones((d.shape[0],), dtype=int)*i
+        for (i, d) in enumerate(datas)
+    ])
+    N = data.shape[0]
+    idx = np.arange(N)
+    np.random.shuffle(idx)
+    data = data[idx]
+    label = label[idx]
+    return data, label
+
+def generate_dataset1():
+    mean1 = (0,0)
+    cov1  = np.array([[124,0],[0,225]])
+    d1 = np.random.multivariate_normal(mean1, cov1, (500,))
+    
+    mean2 = (-35, 40)
+    cov2 = np.array([[32,0],[0,33]])
+    d2 = np.random.multivariate_normal(mean2, cov2, (560,))
+    
+    mean3 = ((30, 50))
+    cov3 = np.array([[44, 0], [0, 20]])
+    d3 = np.random.multivariate_normal(mean3, cov3, (300, ))
+    
+    data,_ = shuffle(d1, d2, d3)
+    
+    
+    return (data,data), 3
+
+def generate_dataset3():
+    from sklearn import cluster, datasets
+    n_samples = 1000
+    noisy_circles = datasets.make_circles(n_samples=n_samples, factor=.5, noise=.05)
+    noisy_moons = datasets.make_moons(n_samples=n_samples, noise=.05)
+    
+    data, _ = shuffle(noisy_circles[0])
+    return (data, data), 2
+def generate_dataset2():
+    mean1 = (0,0)
+    cov1  = np.array([[124,0],[0,225]])
+    d1 = np.random.multivariate_normal(mean1, cov1, (500,))
+    
+    mean2 = (-20, -25)
+    cov2 = np.array([[132,0],[0,33]])
+    d2 = np.random.multivariate_normal(mean2, cov2, (560,))
+    
+    mean3 = ((30, 15))
+    cov3 = np.array([[73, 0], [0, 20]])
+    d3 = np.random.multivariate_normal(mean3, cov3, (300, ))
+    
+    mean4 = ((11, -2))
+    cov4 = np.array([[34,0], [0, 138]])
+    d4 = np.random.multivariate_normal(mean4, cov4, (410, ))
+    
+    data,_ = shuffle(d1, d2, d3, d4)
+    
+    
+    return (data,data), 4
\ No newline at end of file
diff --git a/assignment-3/submission/17307130331/tester_demo.py b/assignment-3/submission/17307130331/tester_demo.py
new file mode 100644
index 0000000000000000000000000000000000000000..19ec0e8091691d4aaaa6b53dbb695fde9e826d89
--- /dev/null
+++ b/assignment-3/submission/17307130331/tester_demo.py
@@ -0,0 +1,117 @@
+import numpy as np
+import sys
+
+from source import KMeans, GaussianMixture
+
+
+def shuffle(*datas):
+    data = np.concatenate(datas)
+    label = np.concatenate([
+        np.ones((d.shape[0],), dtype=int)*i
+        for (i, d) in enumerate(datas)
+    ])
+    N = data.shape[0]
+    idx = np.arange(N)
+    np.random.shuffle(idx)
+    data = data[idx]
+    label = label[idx]
+    return data, label
+
+
+def data_1():
+    mean = (1, 2)
+    cov = np.array([[73, 0], [0, 22]])
+    x = np.random.multivariate_normal(mean, cov, (800,))
+
+    mean = (16, -5)
+    cov = np.array([[21.2, 0], [0, 32.1]])
+    y = np.random.multivariate_normal(mean, cov, (200,))
+
+    mean = (10, 22)
+    cov = np.array([[10, 5], [5, 10]])
+    z = np.random.multivariate_normal(mean, cov, (1000,))
+
+    data, _ = shuffle(x, y, z)
+    return (data, data), 3
+
+
+def data_2():
+    train_data = np.array([
+        [23, 12, 173, 2134],
+        [99, -12, -126, -31],
+        [55, -145, -123, -342],
+    ])
+    return (train_data, train_data), 2
+
+
+def data_3():
+    train_data = np.array([
+        [23],
+        [-2999],
+        [-2955],
+    ])
+    return (train_data, train_data), 2
+
+
+def test_with_n_clusters(data_fuction, algorithm_class):
+    (train_data, test_data), n_clusters = data_fuction()
+    model = algorithm_class(n_clusters)
+    model.fit(train_data)
+    res = model.predict(test_data)
+    assert len(
+        res.shape) == 1 and res.shape[0] == test_data.shape[0], "shape of result is wrong"
+    return res
+
+
+def testcase_1_1():
+    test_with_n_clusters(data_1, KMeans)
+    return True
+
+
+def testcase_1_2():
+    res = test_with_n_clusters(data_2, KMeans)
+    return res[0] != res[1] and res[1] == res[2]
+
+
+def testcase_2_1():
+    test_with_n_clusters(data_1, GaussianMixture)
+    return True
+
+
+def testcase_2_2():
+    res = test_with_n_clusters(data_3, GaussianMixture)
+    return res[0] != res[1] and res[1] == res[2]
+
+
+def test_all(err_report=False):
+    testcases = [
+        ["KMeans-1", testcase_1_1, 4],
+        ["KMeans-2", testcase_1_2, 4],
+        # ["KMeans-3", testcase_1_3, 4],
+        # ["KMeans-4", testcase_1_4, 4],
+        # ["KMeans-5", testcase_1_5, 4],
+        ["GMM-1", testcase_2_1, 4],
+        ["GMM-2", testcase_2_2, 4],
+        # ["GMM-3", testcase_2_3, 4],
+        # ["GMM-4", testcase_2_4, 4],
+        # ["GMM-5", testcase_2_5, 4],
+    ]
+    sum_score = sum([case[2] for case in testcases])
+    score = 0
+    for case in testcases:
+        try:
+            res = case[2] if case[1]() else 0
+        except Exception as e:
+            if err_report:
+                print("Error [{}] occurs in {}".format(str(e), case[0]))
+            res = 0
+        score += res
+        print("+ {:14} {}/{}".format(case[0], res, case[2]))
+    print("{:16} {}/{}".format("FINAL SCORE", score, sum_score))
+
+
+if __name__ == "__main__":
+    if len(sys.argv) > 1 and sys.argv[1] == "--report":
+        test_all(True)
+    else:
+        test_all()