import matplotlib.pyplot as plt import numpy as np import tensorflow as tf sess=tf.Session() x_vals=np.linspace(0,10,100) y_vals=x_vals+np.random.normal(0,1,100) #装换成矩阵 reshape成 (None,1) x_vals_column=np.transpose(np.matrix(x_vals)) #用来合并 100个元素 （100，2）*（2,1）=(100,1) ==> 若为一个元素 （1，2）*（2,1）=(1,1) 符合运算形式 [2,1]*[[1],[2]]=[4] 即为 kx+b=y =>(x,1)*(k,b)=y==>A*x=b ones_column=np.transpose(np.matrix(np.repeat(1,100))) #tuple 参数 A=np.column_stack((x_vals_column,ones_column)) b=np.transpose(np.matrix(y_vals)) #reshape A_tensor=tf.constant(A) b_tensor=tf.constant(b) #(2,2) 简化运算 *转置矩阵 降维 #A*x=b ==> A^T*A*x=A^t*b ==> tA_A * x = A^T*b tA_A=tf.matmul(tf.transpose(A_tensor),A_tensor) #A*x=b => cholseky() 分解矩阵 tA_A=LL' ==> LL'* x=A^T*b L=tf.cholesky(tA_A) #得到A^t*b tA_b=tf.matmul(tf.transpose(A_tensor),b_tensor) # L'* x solve1=tf.matrix_solve(L,tA_b) #x solve2=tf.matrix_solve(tf.transpose(L),solve1) solution_eval=sess.run(solve2) #获取最终结果 slope=solution_eval[0][0] y_intercept=solution_eval[1][0] print(str(slope)) print(str(y_intercept)) best_fit=[] for i in x_vals: best_fit.append(slope*i+y_intercept) plt.plot(x_vals,y_vals,'o') plt.plot(x_vals,best_fit,'r-') plt.show()
