Question

我是机器学习的新手，试图了解tensorflow 2.0的工作原理，但是在Internet上找不到使用新API的全面资源。因此，我已经开始自己探索它了，这是我发现的一个大问题-我该如何训练渴望执行的模型？那是我编写的实现SVM线性模型的代码（使用Tensorflow食谱中的示例），现在我只是不知道接下来要做什么以训练该模型。谢谢您的帮助。

import tensorflow as tf
import numpy as np
import random


alpha_ = 0.1


number_of_predictors = 20
number_of_samples = 100
number_of_predictions = 3


#
# set up dataset
#


predictors = [np.random.normal(size = number_of_predictors).astype('f') for i in range(number_of_samples)]
labels = [[random.choice([-1., 1.]) for i in range(number_of_predictions)] for i in range(number_of_samples)]


#
# set up model
#


slope = tf.Variable(tf.random.normal(shape=[number_of_predictors, number_of_predictions]))
intercept = tf.Variable(tf.random.normal(shape=[1, number_of_predictions]))


# model prediction
predictions = tf.subtract(tf.matmul(predictors, slope), intercept)


# regularization
l2_norm = tf.reduce_sum(tf.square(slope))
alpha = tf.constant([alpha_])


# assemble model prediction and regularization
classification_term = tf.reduce_mean(tf.maximum(0., tf.subtract(1., tf.multiply(predictions, labels))))
loss = tf.add(classification_term, tf.multiply(l2_norm, alpha))


prediction = tf.sign(predictions)
print(f"Predictions: {predictions}")
accuracy = tf.reduce_mean(tf.cast(tf.equal(predictions, labels), dtype = tf.float32))
print(f"Accuracy: {accuracy}")

在Tensorflow 2.0中进行培训，渴望在没有keras的情况下执行

0 个答案: