Question

运行下面的代码后，我获得了精度值，并且可以获得所有Ws和bs的值。我的问题是，将来如何使用输出对事物进行分类？以及如何保存模型？

#########################################
################ SETUP ##################

training_epochs = 500
n_neurons_in_h1 = 60
n_neurons_in_h2 = 60
learning_rate = 0.01

n_features = 3
n_classes = 3

X = tf.placeholder(tf.float32, [None, n_features], name='features')
Y = tf.placeholder(tf.float32, [None, n_classes], name='labels')

W1 = tf.Variable(tf.truncated_normal([n_features, n_neurons_in_h1], mean=0, stddev=1 / np.sqrt(n_features)), name='weights1')
b1 = tf.Variable(tf.truncated_normal([n_neurons_in_h1],mean=0, stddev=1 / np.sqrt(n_features)), name='biases1')
y1 = tf.nn.tanh((tf.matmul(X, W1)+b1), name='activationLayer1')

W2 = tf.Variable(tf.random_normal([n_neurons_in_h1, n_neurons_in_h2],mean=0,stddev=1/np.sqrt(n_features)),name='weights2')
b2 = tf.Variable(tf.random_normal([n_neurons_in_h2],mean=0,stddev=1/np.sqrt(n_features)),name='biases2')
y2 = tf.nn.sigmoid((tf.matmul(y1,W2)+b2),name='activationLayer2')

Wo = tf.Variable(tf.random_normal([n_neurons_in_h2, n_classes], mean=0, stddev=1/np.sqrt(n_features)), name='weightsOut')
bo = tf.Variable(tf.random_normal([n_classes], mean=0, stddev=1/np.sqrt(n_features)), name='biasesOut')
a = tf.nn.softmax((tf.matmul(y2, Wo) + bo), name='activationOutputLayer')

cross_entropy = tf.reduce_mean(-tf.reduce_sum(Y * tf.log(a),reduction_indices=[1]))
train_step = tf.train.GradientDescentOptimizer(learning_rate).minimize(cross_entropy)

correct_prediction = tf.equal(tf.argmax(a, 1), tf.argmax(Y, 1))
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32), name="Accuracy")

#########################################
############### GET DATA ################

（为简洁起见，已删除）

数据的格式为[var1，var2，var3，标签]。被分成培训的80％和测试的20％。

#########################################
################ RUN ####################

# initialization of all variables
initial = tf.global_variables_initializer()

#creating a session
with tf.Session() as sess:
    sess.run(initial)

    # training loop over the number of epoches
    batchsize=10
    for epoch in range(training_epochs):
        for i in range(len(tr_features)):

            start=i
            end=i+batchsize
            x_batch=tr_features[start:end]
            y_batch=tr_labels[start:end]

            # feeding training data/examples
            sess.run(train_step, feed_dict={X:x_batch , Y:y_batch})
            i+=batchsize
        # feeding testing data to determine model accuracy
        y_pred = sess.run(tf.argmax(a, 1), feed_dict={X: ts_features})
        y_true = sess.run(tf.argmax(ts_labels, 1))
        acc = sess.run(accuracy, feed_dict={X: ts_features, Y: ts_labels})

        # print accuracy for each epoch
        print('epoch',epoch, acc)
        print ('---------------')
        print(y_pred, y_true)

Answer 1

您可以使用tf.train.Saver()类保存模型（权重+附加值）。此类的save方法将保存所有内容，而restore方法将加载由save方法创建的文件，因此您无需重新训练即可进行预测。

查看文档：{{3}}

我如何使用TensorFlow神经网络输出

1 个答案: