3-Tensorflow-demo_00-tf和numpy区别

# -*- coding: utf-8 -*-
import tensorflow as tf
import numpy as np

"""
演示numpy 和tensorflow的区别
"""

np.random.seed(42)

x_data = np.random.rand(100).astype(np.float32)
#通过本函数可以返回一个或一组服从“0~1”均匀分布的随机样本值。
y_data = x_data * 0.1 +0.3
print('x_data.shape:',x_data.shape)
print('y_data:',y_data)

#一，构建模型图
with tf.Graph().as_default():
    # 1.构建变量
    w = tf.Variable(initial_value=tf.random_uniform([1],-1.0,1.0))##生成一维向量，大小在-1到1之间
    b = tf.Variable(initial_value=tf.zeros([1]))#shape代表形状，也就是1纬的还是2纬的还是n纬的数组。

    #2.执行正向传播，获得预测值
    y_pred = x_data * w +b

    #3.构建损失函数
    loss = tf.reduce_mean(tf.square(y_pred - y_data))

    #4.构建优化器
    optimizer = tf.train.GradientDescentOptimizer(learning_rate=0.01)#执行梯度下降优化器
    train_opt = optimizer.minimize(loss=loss)
    print(w,b,loss,train_opt)

    # 二、构建会话
    with tf.Session() as sess:
        #1.变量初始化
        sess.run(tf.global_variables_initializer())
        #2.构建迭代循环
        for e in range(1,201):
            #3.执行模型训练
            _,loss_ = sess.run([train_opt,loss])
            if e % 20 == 0:
                print('Epoch:{} - Train Loss:{:.5f}'.format(e,loss_))
                print(sess.run(w),sess.run(b))

x_data.shape: (100,)
y_data: [0.33745402 0.39507145 0.3731994  0.35986587 0.3156019  0.31559947
 0.30580837 0.38661763 0.3601115  0.3708073  0.30205846 0.396991
 0.38324428 0.32123393 0.3181825  0.31834045 0.33042425 0.35247564
 0.3431945  0.32912293 0.3611853  0.3139494  0.32921448 0.3366362
 0.345607   0.37851763 0.3199674  0.35142344 0.35924146 0.30464506
 0.3607545  0.31705242 0.30650517 0.39488858 0.39656323 0.38083977
 0.33046138 0.30976722 0.3684233  0.34401527 0.31220382 0.3495177
 0.30343887 0.39093205 0.32587802 0.36625224 0.33117113 0.35200682
 0.35467103 0.31848547 0.39695847 0.3775133  0.39394993 0.38948274
 0.35979    0.39218745 0.30884928 0.3195983  0.30452275 0.33253303
 0.33886775 0.3271349  0.38287377 0.33567536 0.32809347 0.35426962
 0.31409243 0.3802197  0.30745506 0.3986887  0.3772245  0.31987157
 0.30055222 0.38154617 0.37068576 0.37290072 0.37712705 0.3074045
 0.33584657 0.31158692 0.38631037 0.3623298  0.33308983 0.30635583
 0.33109826 0.33251834 0.37296063 0.36375576 0.3887213  0.3472215
 0.31195945 0.37132448 0.37607852 0.35612774 0.37709674 0.34937957
 0.3522733  0.34275413 0.3025419  0.31078917]
<tf.Variable 'Variable:0' shape=(1,) dtype=float32_ref> <tf.Variable 'Variable_1:0' shape=(1,) dtype=float32_ref> Tensor("Mean:0", shape=(), dtype=float32) name: "GradientDescent"
op: "NoOp"
input: "^GradientDescent/update_Variable/ApplyGradientDescent"
input: "^GradientDescent/update_Variable_1/ApplyGradientDescent"

Epoch:20 - Train Loss:0.19688
[-0.5349532] [0.20760678]
Epoch:40 - Train Loss:0.08353
[-0.45615944] [0.33085623]
Epoch:60 - Train Loss:0.04132
[-0.40364257] [0.40308142]
Epoch:80 - Train Loss:0.02525
[-0.3671751] [0.44446975]
Epoch:100 - Train Loss:0.01880
[-0.34055865] [0.46724656]
Epoch:120 - Train Loss:0.01591
[-0.32003522] [0.47881463]
Epoch:140 - Train Loss:0.01435
[-0.30332527] [0.48365596]
Epoch:160 - Train Loss:0.01331
[-0.28904456] [0.48448348]
Epoch:180 - Train Loss:0.01248
[-0.27635142] [0.48293874]
Epoch:200 - Train Loss:0.01176
[-0.26473287] [0.4800143]

Process finished with exit code 0