简单神经网络解决二分类问题示例（Tensorflow）（自定义损失函数）

简单神经网络解决二分类问题示例（Tensorflow）（自定义损失函数）

---

在预测商品销量时，如果预测多了，商家损失的是生产商品的成本;而如果预测少了，损失的则是商品的利润。针对这种模型可以自己定义损失函数。

自定义损失函数为：

$$Loss(y,y')=\sum_{i=1}^{n}f(y_{_i},y_{i}^{'}), f(x,y)=\left\{\begin{matrix} & a(x-y);x>y & \\ & b(y-x);x<=y& \end{matrix}\right.$$

可使用如下代码实现：

loss_less = 10
loss_more = 1
loss = tf.reduce_sum(tf.where(tf.greater(y, y_),
                        (y - y_) * loss_more,
                        (y_ - y) * loss_less))

整体代码如下：

#coding:utf-8
import tensorflow as tf
from numpy.random import RandomState

batch_size = 8

x = tf.placeholder(tf.float32, shape = (None, 2), name = 'x-input')
y_ = tf.placeholder(tf.float32, shape = (None, 1), name = 'y-input')

w1 = tf.Variable(tf.random_normal([2,1],stddev = 1, seed = 1))
y = tf.matmul(x, w1)

loss_less = 10
loss_more = 1
loss = tf.reduce_sum(tf.where(tf.greater(y, y_),
                        (y - y_) * loss_more,
                        (y_ - y) * loss_less))
train_step = tf.train.AdamOptimizer(0.001).minimize(loss)

rdm  = RandomState(1)
dataset_size = 128
X = rdm.rand(dataset_size, 2)
Y = [[x1 + x2 + rdm.rand()/10.0-0.05] for (x1, x2) in X]

with tf.Session() as sess:
    init_op = tf.initialize_all_variables()
    sess.run(init_op)
    STEPS = 5000
    for i in range(STEPS):
        start = (i * batch_size) % dataset_size
        end = min(start+batch_size, dataset_size)
        sess.run(train_step, feed_dict={x : X[start:end], y_ : Y[start:end]})
        print(sess.run(w1))