多分类DNN | MLP多层感知机的原理及代码实现----tensorflow | DnnClassifier

1、多层感知机原理MLP  |  DNN

         前向传播

         后向传播

         激活函数

         梯度下降

2、代码实现对鸢尾花数据分类

        主程序：分类结果最终可到达1.0，完全正确       

import tensorflow as tf
from data_iris import train_data_process

# Parameters
learning_rate = 0.001
training_epochs = 1500
batch_size = 60
display_step = 1

# Network Parameters
n_hidden_1 = 20 # 1st layer number of neurons
n_hidden_2 = 20 # 2nd layer number of neurons
n_input = 4 # MNIST data input (img shape: 28*28)
n_classes = 3 # MNIST total classes (0-9 digits)

# tf Graph input
X = tf.placeholder("float", [None, n_input])
Y = tf.placeholder("float", [None, n_classes])

# Store layers weight & bias
weights = {
    'h1': tf.Variable(tf.random_normal([n_input, n_hidden_1])),
    'h2': tf.Variable(tf.random_normal([n_hidden_1, n_hidden_2])),
    'out': tf.Variable(tf.random_normal([n_hidden_2, n_classes]))
}
biases = {
    'b1': tf.Variable(tf.random_normal([n_hidden_1])),
    'b2': tf.Variable(tf.random_normal([n_hidden_2])),
    'out': tf.Variable(tf.random_normal([n_classes]))
}


# 创建模型
def multilayer_perceptron(x):
    # Hidden fully connected layer with 256 neurons
    layer_1 = tf.add(tf.matmul(x, weights['h1']), biases['b1'])
    # Hidden fully connected layer with 256 neurons
    layer_2 = tf.add(tf.matmul(layer_1, weights['h2']), biases['b2'])
    # Output fully connected layer with a neuron for each class
    out_layer = tf.matmul(layer_2, weights['out']) + biases['out']
    return out_layer

logits = multilayer_perceptron(X)

# Define loss and optimizer
loss_op = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits=logits, labels=Y))
optimizer = tf.train.AdamOptimizer(learning_rate=learning_rate)
train_op = optimizer.minimize(loss_op)

# 初始化
init = tf.global_variables_initializer()

with tf.Session() as sess:
    sess.run(init)

    # 训练循环
    for epoch in range(training_epochs):
        avg_cost = 0.
        total_batch = int(120/batch_size)
        # 遍历所有批次
        for i in range(total_batch):
            batch_x, batch_y, _ = train_data_process(batch_size)
            # 运行优化器op  和  损失op
            _, c = sess.run([train_op, loss_op], feed_dict={X: batch_x, Y: batch_y})
            # 计算平均损失
            avg_cost += c / total_batch
        # Display logs per epoch step
        if epoch % display_step == 0:
            print("Epoch:", '%04d' % (epoch+1), "cost={:.9f}".format(avg_cost))
    print("Optimization Finished!")

    # 对输出归一化，归一化也有可能是数据太小，不一定都是很大
    pred = tf.nn.softmax(logits)  # Apply softmax to logits
    correct_prediction = tf.equal(tf.argmax(pred, 1), tf.argmax(Y, 1))
    # 计算精确率，这里没有开启一个回话，使用的是eval取值
    accuracy = tf.reduce_mean(tf.cast(correct_prediction, "float"))

    test_x, test_y, _ = train_data_process(batch_size)
    print("Accuracy:", accuracy.eval({X: test_x, Y: test_y}))

          数据导入程序：

import pandas as pd
import pdb
import numpy as np
from sklearn.preprocessing import LabelBinarizer


# 抽取数据，每次要抽取20个，batch_size = 20+1
# 返回特征数据 标签 标签名称
# 这里小心抽出的样本标签没有涵盖全部情况，会导致onehot编码出现降维的情况，就不修改了，自己看着改吧
def train_data_process(batch_size):
	# 读入数据
	train = pd.read_csv('./iris_training.csv').dropna().drop(index=[0]).sample(batch_size+1)

	# test = pd.read_csv('./iris_test.csv').dropna()
	# print(list(train.iloc[0, :]))

	batch_train = np.array(train.iloc[1:, 0:4]).tolist()

	encoder = LabelBinarizer()
	batch_label = encoder.fit_transform(np.array(train.iloc[1:, 4]).tolist())

	label_name = list(train.iloc[0:1,2:5])
	# pdb.set_trace()

	# 返回特征数据 标签 标签名称
	return batch_train, batch_label, label_name 


# 测试数据
def test_data_process(batch_size):
	test = pd.read_csv('./iris_test.csv').dropna().drop(index=[0]).sample(batch_size+1)
	batch_test = np.array(test.iloc[1:, 0:4]).tolist()

	encoder = LabelBinarizer()
	batch_label = encoder.fit_transform(np.array(test.iloc[1:, 4]).tolist())

	return batch_test, batch_label


if __name__ == '__main__':
	print(train_data_process(21))