在线矩阵微积分工具，可以生成 Python/Latex 代码哦！

社区在线技术交流群

https://bbs.mlqi.org

（大家多去逛逛哈）

今天给大家介绍一个实用的在线小工具，矩阵微积分在线生成 Python/Latex 代码。

链接：

http://www.matrixcalculus.org/matrixCalculus

Python代码如下：

"""
d/dx x'*A*x + c*sin(y)'*x = 2*A*x+c*sin(y)
A is a matrix
c is a scalar
y is a vector
x is a vector
"""

from __future__ import division, print_function, absolute_import
   
import numpy as np
   
def fAndG(A, c, x, y):
   assert(type(A) == np.ndarray)
   dim = A.shape
   assert(len(dim) == 2)
   A_rows = dim[0]
   A_cols = dim[1]
   if type(c) == np.ndarray:
       dim = c.shape
       assert(dim == (1, ))
   assert(type(x) == np.ndarray)
   dim = x.shape
   assert(len(dim) == 1)
   x_rows = dim[0]
   assert(type(y) == np.ndarray)
   dim = y.shape
   assert(len(dim) == 1)
   y_rows = dim[0]
   assert(A_cols == x_rows == y_rows == A_rows)

   t_0 = np.dot(A, x)
   t_1 = np.sin(y)
   functionValue = (np.dot(x, t_0) + (c * np.dot(t_1, x)))
   gradient = ((2 * t_0) + (c * t_1))

   return functionValue, gradient

def generateRandomData():
   A = np.random.randn(3, 3)
   c = np.random.randn(1)
   x = np.random.randn(3)
   y = np.random.randn(3)
   return A, c, x, y

if __name__ == '__main__':
   A, c, x, y = generateRandomData()
   functionValue, gradient = fAndG(A, c, x, y)
   print('functionValue = ', functionValue)
   print('gradient = ', gradient)

Latex代码如下：

documentclass[12pt]{article}
usepackage{amsmath,amsthm,amssymb}
begin{document}
function:
begin{align*}
 f = x^top cdot Acdot x+ccdot sin(y)^top cdot x
end{align*}

gradient:
begin{align*}
 frac{partial f}{partial x} = 2cdot Acdot x+ccdot sin(y)
end{align*}
end{document}