当前位置：首页 > article >正文

tf.Keras (tf-1.15)使用记录1-基础模型创建的两种方法

article 2025/2/3 4:39:28

tf.keras有两种神经网络的创建方法，分别是基于sequential的方式创建和基于functional的方式创建。其中基于functional的方式创建更为灵活，更容易自定义网络中的细节；基于sequential的方式创建则更加简单。

1）基于sequential方式处理pd数据示例：更简单

不需要定义和处理输入输出，以及内部的数据流。

import numpy as np
import pandas as pd
from sklearn.model_selection import train_test_split
import tensorflow as tf

# 生成数据并转化为pandas
X = np.random.randn(1000, 2)
Y = (X[:, 0] + X[:, 1] > 0).astype(int)
df = pd.DataFrame(X, columns=['feature1', 'feature2'])
df['target'] = Y

# 分割数据
X_train, X_test, y_train, y_test = train_test_split(df[['feature1', 'feature2']], df['target'], test_size=0.2, random_state=42)

# tf.keras模型定义
# 构建Sequential模型
model = tf.keras.Sequential([
    tf.keras.layers.Dense(64, activation='relu', input_shape=(2,)),
    tf.keras.layers.BatchNormalization(),
    tf.keras.layers.Dropout(0.5),
    tf.keras.layers.Dense(32, activation='relu'),
    tf.keras.layers.BatchNormalization(),
    tf.keras.layers.Dropout(0.3),
    tf.keras.layers.Dense(1, activation='sigmoid')
])

# 编译模型
model.compile(optimizer='adam', loss='binary_crossentropy', metrics=['accuracy'])

# 训练模型
model.fit(X_train, y_train, epochs=10, batch_size=32, validation_split=0.1, verbose=1)

# 预测
preds = model.predict(X_test)

2）基于functional方式处理pd数据示例：更灵活

import numpy as np
import pandas as pd
from sklearn.model_selection import train_test_split
import tensorflow as tf

# 生成数据并转化为pandas
X = np.random.randn(1000, 2)
Y = (X[:, 0] + X[:, 1] > 0).astype(int)
df = pd.DataFrame(X, columns=['feature1', 'feature2'])
df['target'] = Y

# 分割数据
X_train, X_test, y_train, y_test = train_test_split(df[['feature1', 'feature2']], df['target'], test_size=0.2, random_state=42)

# tf.keras模型定义
# 构建Functional API模型
def get_my_model(input_dim):
    inputs = tf.keras.Input(shape=(input_dim,))
    x = tf.keras.layers.Dense(64, activation='relu')(inputs)
    x = tf.keras.layers.BatchNormalization()(x)
    x = tf.keras.layers.Dropout(0.3)(x)
    x = tf.keras.layers.Dense(32, activation='relu')(x)
    x = tf.keras.layers.BatchNormalization()(x)
    x = tf.keras.layers.Dropout(0.3)(x)    
    outputs = tf.keras.layers.Dense(1, activation='sigmoid')(x)
    model = tf.keras.Model(inputs=inputs, outputs=outputs)
    model.compile(optimizer='adam', loss='binary_crossentropy', metrics=['accuracy'])
    return model

# 编译模型
model = get_my_model(2)

# 训练模型
model.fit(X_train, y_train, epochs=10, batch_size=32, validation_split=0.1, verbose=1)

# 预测
preds = model.predict(X_test)