MLP实现minist数据集分类任务

1. 数据集

minist手写体数字数据集

2. 代码

?

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127

''' Description: Author: zhangyh Date: 2024-05-04 15:21:49 LastEditTime: 2024-05-04 22:36:26 LastEditors: zhangyh '''   import numpy as np   class MlpClassifier:        def __init__(self, input_size, hidden_size1, hidden_size2, output_size, learning_rate=0.01):         self.input_size = input_size         self.hidden_size1 = hidden_size1         self.hidden_size2 = hidden_size2         self.output_size = output_size         self.learning_rate = learning_rate           self.W1 = np.random.randn(input_size, hidden_size1) * 0.01         self.b1 = np.zeros((1, hidden_size1))         self.W2 = np.random.randn(hidden_size1, hidden_size2) * 0.01         self.b2 = np.zeros((1, hidden_size2))         self.W3 = np.random.randn(hidden_size2, output_size) * 0.01         self.b3 = np.zeros((1, output_size))           def softmax(self, x):         exps = np.exp(x - np.max(x, axis=1, keepdims=True))         return exps / np.sum(exps, axis=1, keepdims=True)           def relu(self, x):         return np.maximum(x, 0)           def relu_derivative(self, x):         return np.where(x > 0, 1, 0)           def cross_entropy_loss(self, y_true, y_pred):         m = y_true.shape[0]         return -np.sum(y_true * np.log(y_pred + 1e-8)) / m           def forward(self, X):         self.Z1 = np.dot(X, self.W1) + self.b1         self.A1 = self.relu(self.Z1)         self.Z2 = np.dot(self.A1, self.W2) + self.b2         self.A2 = self.relu(self.Z2)         self.Z3 = np.dot(self.A2, self.W3) + self.b3         self.A3 = self.softmax(self.Z3)         return self.A3           def backward(self, X, y):         m = X.shape[0]         dZ3 = self.A3 - y         dW3 = np.dot(self.A2.T, dZ3) / m         db3 = np.sum(dZ3, axis=0, keepdims=True) / m         dA2 = np.dot(dZ3, self.W3.T)         dZ2 = dA2 * self.relu_derivative(self.Z2)         dW2 = np.dot(self.A1.T, dZ2) / m         db2 = np.sum(dZ2, axis=0, keepdims=True) / m         dA1 = np.dot(dZ2, self.W2.T)         dZ1 = dA1 * self.relu_derivative(self.Z1)         dW1 = np.dot(X.T, dZ1) / m         db1 = np.sum(dZ1, axis=0, keepdims=True) / m                   # Update weights and biases         self.W3 -= self.learning_rate * dW3         self.b3 -= self.learning_rate * db3         self.W2 -= self.learning_rate * dW2         self.b2 -= self.learning_rate * db2         self.W1 -= self.learning_rate * dW1         self.b1 -= self.learning_rate * db1       # 计算精确度     def accuracy(self, y_pred, y):         predictions = np.argmax(y_pred, axis=1)         correct_predictions = np.sum(predictions == np.argmax(y, axis=1))            return correct_predictions / y.shape[0]           def train(self, X, y, epochs=100, batch_size=64):         print('Training...')            m = X.shape[0]         for epoch in range(epochs):             for i in range(0, m, batch_size):                 X_batch = X[i:i+batch_size]                 y_batch = y[i:i+batch_size]                                   # Forward propagation                 y_pred = self.forward(X_batch)                                   # Backward propagation                 self.backward(X_batch, y_batch)                           if (epoch+1) % 10 == 0:                 loss = self.cross_entropy_loss(y, self.forward(X))                 acc = self.accuracy(y_pred, y_batch)                 print(f'Epoch {epoch+1}/{epochs}, Loss: {loss}, Training-Accuracy: {acc}')         def test(self, X, y):         print('Testing...')         y_pred = self.forward(X)         acc = self.accuracy(y_pred, y)            return acc     if __name__ == '__main__':        import tensorflow as tf       # 加载MNIST数据集     (X_train, y_train), (X_test, y_test) = tf.keras.datasets.mnist.load_data()       # 将图像转换为向量形式     X_train = X_train.reshape(X_train.shape[0], -1) / 255.0     X_test = X_test.reshape(X_test.shape[0], -1) / 255.0     # 将标签进行 one-hot 编码     num_classes = 10     y_train = tf.keras.utils.to_categorical(y_train, num_classes)     y_test = tf.keras.utils.to_categorical(y_test, num_classes)       # 打印转换后的结果     # 训练集维度: (60000, 784) (60000, 10)     # 测试集维度: (10000, 784) (10000, 10)     model = MlpClassifier(784, 128, 128, 10)       model.train(X_train, y_train)         test_acc = model.test(X_test, y_test)      print(f'Test-Accuracy: {test_acc}')

　　

3. 运行结果

?

1 2 3 4 5 6 7 8 9 10 11 12 13

Training... Epoch 10/100, Loss: 0.3617846299623725, Training-Accuracy: 0.9375 Epoch 20/100, Loss: 0.1946690996652946, Training-Accuracy: 1.0 Epoch 30/100, Loss: 0.13053815227522408, Training-Accuracy: 1.0 Epoch 40/100, Loss: 0.09467908427578901, Training-Accuracy: 1.0 Epoch 50/100, Loss: 0.07120217251250453, Training-Accuracy: 1.0 Epoch 60/100, Loss: 0.055233734086591456, Training-Accuracy: 1.0 Epoch 70/100, Loss: 0.04369171830999816, Training-Accuracy: 1.0 Epoch 80/100, Loss: 0.03469674775956587, Training-Accuracy: 1.0 Epoch 90/100, Loss: 0.027861857647949812, Training-Accuracy: 1.0 Epoch 100/100, Loss: 0.0225212692988995, Training-Accuracy: 1.0 Testing... Test-Accuracy: 0.9775