DAY 41 简单CNN

最新推荐文章于 2026-01-08 09:31:05 发布

原创最新推荐文章于 2026-01-08 09:31:05 发布 · 173 阅读

0 ·

CC 4.0 BY-SA版权

文章标签：

#cnn #人工智能 #神经网络

Python入门（坚持）专栏收录该内容

52 篇文章

订阅专栏

知识回顾

数据增强
卷积神经网络定义的写法
batch归一化：调整一个批次的分布，常用与图像数据
特征图：只有卷积操作输出的才叫特征图
调度器：直接修改基础学习率

卷积操作常见流程如下：

1. 输入 → 卷积层 → Batch归一化层（可选） → 池化层 → 激活函数 → 下一层

Flatten -> Dense (with Dropout，可选) -> Dense (Output)

作业：尝试手动修改下不同的调度器和CNN的结构，观察训练的差异。

import numpy as np
from tensorflow import keras
from tensorflow.keras import layers

# 加载和预处理数据
(x_train, y_train), (x_test, y_test) = keras.datasets.mnist.load_data()
x_train = x_train.reshape(-1, 28, 28, 1).astype("float32") / 255.0
x_test = x_test.reshape(-1, 28, 28, 1).astype("float32") / 255.0
y_train = keras.utils.to_categorical(y_train, 10)
y_test = keras.utils.to_categorical(y_test, 10)

# 定义简单的 CNN 模型
def simple_cnn():
    model = keras.Sequential([
        layers.Conv2D(16, (3, 3), activation='relu', input_shape=(28, 28, 1)),
        layers.MaxPooling2D((2, 2)),
        layers.Flatten(),
        layers.Dense(128, activation='relu'),
        layers.Dense(10, activation='softmax')
    ])
    return model

# 定义复杂的 CNN 模型
def complex_cnn():
    model = keras.Sequential([
        layers.Conv2D(32, (3, 3), activation='relu', input_shape=(28, 28, 1)),
        layers.MaxPooling2D((2, 2)),
        layers.Conv2D(64, (3, 3), activation='relu'),
        layers.MaxPooling2D((2, 2)),
        layers.Flatten(),
        layers.Dense(256, activation='relu'),
        layers.Dense(128, activation='relu'),
        layers.Dense(10, activation='softmax')
    ])
    return model

# 定义不同的优化器
optimizers = {
    'SGD': keras.optimizers.SGD(learning_rate=0.01),
    'Adam': keras.optimizers.Adam(learning_rate=0.001)
}

# 训练不同的模型和优化器组合
epochs = 5
batch_size = 64

for model_name, model_fn in [('Simple CNN', simple_cnn), ('Complex CNN', complex_cnn)]:
    for optimizer_name, optimizer in optimizers.items():
        model = model_fn()
        model.compile(optimizer=optimizer, loss='categorical_crossentropy', metrics=['accuracy'])
        print(f"Training {model_name} with {optimizer_name} optimizer:")
        history = model.fit(x_train, y_train, epochs=epochs, batch_size=batch_size, validation_data=(x_test, y_test))

        train_loss = history.history['loss']
        train_acc = history.history['accuracy']
        val_loss = history.history['val_loss']
        val_acc = history.history['val_accuracy']

        print(f"Training Loss: {train_loss}")
        print(f"Training Accuracy: {train_acc}")
        print(f"Validation Loss: {val_loss}")
        print(f"Validation Accuracy: {val_acc}")