TensorFlow Addons 优化器：ConditionalGradient

在 TensorFlow.org 上查看 在 Google Colab 中运行 在 GitHub 上查看源代码 下载笔记本

概述

此笔记本将演示如何使用 Addons 软件包中的条件梯度优化器。

ConditionalGradient

由于潜在的正则化效果，约束神经网络的参数已被证明对训练有益。通常，参数通过软惩罚（从不保证约束满足）或通过投影运算（计算资源消耗大）进行约束。另一方面，条件梯度 (CG) 优化器可严格执行约束，而无需消耗资源的投影步骤。它通过最大程度减小约束集中目标的线性逼近来工作。在此笔记本中，我们通过 MNIST 数据集上的 CG 优化器演示弗罗宾尼斯范数约束的应用。CG 现在可以作为 Tensorflow API 提供。有关优化器的更多详细信息，请参阅 https://arxiv.org/pdf/1803.06453.pdf

设置

pip install -U tensorflow-addons

import tensorflow as tf
import tensorflow_addons as tfa
from matplotlib import pyplot as plt

# Hyperparameters
batch_size=64
epochs=10

构建模型

model_1 = tf.keras.Sequential([
    tf.keras.layers.Dense(64, input_shape=(784,), activation='relu', name='dense_1'),
    tf.keras.layers.Dense(64, activation='relu', name='dense_2'),
    tf.keras.layers.Dense(10, activation='softmax', name='predictions'),
])

准备数据

# Load MNIST dataset as NumPy arrays
dataset = {}
num_validation = 10000
(x_train, y_train), (x_test, y_test) = tf.keras.datasets.mnist.load_data()

# Preprocess the data
x_train = x_train.reshape(-1, 784).astype('float32') / 255
x_test = x_test.reshape(-1, 784).astype('float32') / 255

Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/mnist.npz
11493376/11490434 [==============================] - 0s 0us/step
11501568/11490434 [==============================] - 0s 0us/step

定义自定义回调函数

def frobenius_norm(m):
    """This function is to calculate the frobenius norm of the matrix of all
    layer's weight.

    Args:
        m: is a list of weights param for each layers.
    """
    total_reduce_sum = 0
    for i in range(len(m)):
        total_reduce_sum = total_reduce_sum + tf.math.reduce_sum(m[i]**2)
    norm = total_reduce_sum**0.5
    return norm

CG_frobenius_norm_of_weight = []
CG_get_weight_norm = tf.keras.callbacks.LambdaCallback(
    on_epoch_end=lambda batch, logs: CG_frobenius_norm_of_weight.append(
        frobenius_norm(model_1.trainable_weights).numpy()))

训练和评估：使用 CG 作为优化器

只需用新的 TFA 优化器替换典型的 Keras 优化器 

# Compile the model
model_1.compile(
    optimizer=tfa.optimizers.ConditionalGradient(
        learning_rate=0.99949, lambda_=203),  # Utilize TFA optimizer
    loss=tf.keras.losses.SparseCategoricalCrossentropy(),
    metrics=['accuracy'])

history_cg = model_1.fit(
    x_train,
    y_train,
    batch_size=batch_size,
    validation_data=(x_test, y_test),
    epochs=epochs,
    callbacks=[CG_get_weight_norm])

Epoch 1/10
938/938 [==============================] - 3s 3ms/step - loss: 0.3901 - accuracy: 0.8826 - val_loss: 0.2070 - val_accuracy: 0.9375
Epoch 2/10
938/938 [==============================] - 2s 3ms/step - loss: 0.1871 - accuracy: 0.9442 - val_loss: 0.1567 - val_accuracy: 0.9507
Epoch 3/10
938/938 [==============================] - 2s 3ms/step - loss: 0.1471 - accuracy: 0.9561 - val_loss: 0.1253 - val_accuracy: 0.9635
Epoch 4/10
938/938 [==============================] - 2s 3ms/step - loss: 0.1282 - accuracy: 0.9616 - val_loss: 0.1526 - val_accuracy: 0.9493
Epoch 5/10
938/938 [==============================] - 2s 3ms/step - loss: 0.1181 - accuracy: 0.9644 - val_loss: 0.2061 - val_accuracy: 0.9313
Epoch 6/10
938/938 [==============================] - 2s 3ms/step - loss: 0.1147 - accuracy: 0.9659 - val_loss: 0.1480 - val_accuracy: 0.9559
Epoch 7/10
938/938 [==============================] - 2s 3ms/step - loss: 0.1084 - accuracy: 0.9668 - val_loss: 0.1183 - val_accuracy: 0.9663
Epoch 8/10
938/938 [==============================] - 2s 3ms/step - loss: 0.1051 - accuracy: 0.9682 - val_loss: 0.1599 - val_accuracy: 0.9471
Epoch 9/10
938/938 [==============================] - 2s 2ms/step - loss: 0.1023 - accuracy: 0.9689 - val_loss: 0.1100 - val_accuracy: 0.9665
Epoch 10/10
938/938 [==============================] - 2s 3ms/step - loss: 0.1000 - accuracy: 0.9701 - val_loss: 0.1202 - val_accuracy: 0.9644

训练和评估：使用 SGD 作为优化器

model_2 = tf.keras.Sequential([
    tf.keras.layers.Dense(64, input_shape=(784,), activation='relu', name='dense_1'),
    tf.keras.layers.Dense(64, activation='relu', name='dense_2'),
    tf.keras.layers.Dense(10, activation='softmax', name='predictions'),
])

SGD_frobenius_norm_of_weight = []
SGD_get_weight_norm = tf.keras.callbacks.LambdaCallback(
    on_epoch_end=lambda batch, logs: SGD_frobenius_norm_of_weight.append(
        frobenius_norm(model_2.trainable_weights).numpy()))

# Compile the model
model_2.compile(
    optimizer=tf.keras.optimizers.SGD(0.01),  # Utilize SGD optimizer
    loss=tf.keras.losses.SparseCategoricalCrossentropy(),
    metrics=['accuracy'])

history_sgd = model_2.fit(
    x_train,
    y_train,
    batch_size=batch_size,
    validation_data=(x_test, y_test),
    epochs=epochs,
    callbacks=[SGD_get_weight_norm])

Epoch 1/10
938/938 [==============================] - 2s 2ms/step - loss: 0.9734 - accuracy: 0.7435 - val_loss: 0.4342 - val_accuracy: 0.8835
Epoch 2/10
938/938 [==============================] - 2s 2ms/step - loss: 0.3880 - accuracy: 0.8919 - val_loss: 0.3311 - val_accuracy: 0.9070
Epoch 3/10
938/938 [==============================] - 2s 2ms/step - loss: 0.3228 - accuracy: 0.9086 - val_loss: 0.2885 - val_accuracy: 0.9182
Epoch 4/10
938/938 [==============================] - 2s 2ms/step - loss: 0.2890 - accuracy: 0.9166 - val_loss: 0.2661 - val_accuracy: 0.9233
Epoch 5/10
938/938 [==============================] - 2s 2ms/step - loss: 0.2650 - accuracy: 0.9238 - val_loss: 0.2466 - val_accuracy: 0.9301
Epoch 6/10
938/938 [==============================] - 2s 2ms/step - loss: 0.2457 - accuracy: 0.9295 - val_loss: 0.2293 - val_accuracy: 0.9350
Epoch 7/10
938/938 [==============================] - 2s 2ms/step - loss: 0.2295 - accuracy: 0.9334 - val_loss: 0.2153 - val_accuracy: 0.9385
Epoch 8/10
938/938 [==============================] - 2s 2ms/step - loss: 0.2153 - accuracy: 0.9377 - val_loss: 0.2031 - val_accuracy: 0.9405
Epoch 9/10
938/938 [==============================] - 2s 2ms/step - loss: 0.2027 - accuracy: 0.9409 - val_loss: 0.1936 - val_accuracy: 0.9436
Epoch 10/10
938/938 [==============================] - 2s 2ms/step - loss: 0.1918 - accuracy: 0.9443 - val_loss: 0.1853 - val_accuracy: 0.9447

权重的弗罗宾尼斯范数：CG 与 SGD

CG 优化器的当前实现基于弗罗宾尼斯范数，并考虑将弗罗宾尼斯范数作为目标函数中的正则化器。因此，我们将 CG 的正则化效果与尚未采用弗罗宾尼斯范数正则化器的 SGD 优化器进行比较。

plt.plot(
    CG_frobenius_norm_of_weight,
    color='r',
    label='CG_frobenius_norm_of_weights')
plt.plot(
    SGD_frobenius_norm_of_weight,
    color='b',
    label='SGD_frobenius_norm_of_weights')
plt.xlabel('Epoch')
plt.ylabel('Frobenius norm of weights')
plt.legend(loc=1)

<matplotlib.legend.Legend at 0x7efde5ac4c50>

png

训练和验证准确率：CG 与 SGD

plt.plot(history_cg.history['accuracy'], color='r', label='CG_train')
plt.plot(history_cg.history['val_accuracy'], color='g', label='CG_test')
plt.plot(history_sgd.history['accuracy'], color='pink', label='SGD_train')
plt.plot(history_sgd.history['val_accuracy'], color='b', label='SGD_test')
plt.xlabel('Epoch')
plt.ylabel('Accuracy')
plt.legend(loc=4)

<matplotlib.legend.Legend at 0x7efde514fe10>

png