Lagrange multiplier method （拉格朗日乘数法）

The Lagrange multiplier method is used to solve the problem that find the extremum of a function $z = f(x,y)$, given $\phi (x,y) = 0$.

simple extremum problem

To solve a simple extremum problem,

$$z = f(x,y)$$
we just need to find the $(x,y)$ where
$$\frac{\partial z}{\partial x} = 0\\ \frac{{\partial z}}{{\partial y}} = 0$$
By solving the simultaneous equations, we could find the (x,y)s. Then plugging in the (x,y)s into the function z = f(x,y), we find the extremum of z.

extremum problem with restricted condition

The problem is finding the extremum of

$$z = f(x,y)$$
where,
$$\phi (x,y) = 0$$
By applying Lagrange multiplier method, we define a new function, as:
$$F(x,y,\lambda) = f(x,y) + \lambda\phi(x,y)$$
Then, take the partial derivatives of $x,y,\lambda$. We get $\frac{\partial F}{\partial x}, \frac{\partial F}{\partial y},\frac{\partial F}{\partial \lambda}$. Make them be $0$.
$$\frac{\partial F}{\partial x} = 0\\ \frac{\partial F}{\partial y} = 0\\ \frac{\partial F}{\partial \lambda} = 0\\ $$
Three equations, three unknow variables. The equation set could be solved. By solving the equation set, we get the $(x,y)$, we care nothing about $\lambda$.
Then, plugging the $(x,y)$, just solve out, in the function $z = f(x,y)$, we find the extremum of $z = f(x,y)$ with restricted condition $\phi(x,y) = 0$

explanation

as $\phi(x,y)$ always be $0$, $F(x,y,\lambda)$ will always be equal to $f(x,y)$.

$$F(x,y,\lambda) = f(x,y) + \lambda\phi(x,y)\\ \phi(x,y) = 0\\ \Downarrow\\ F(x,y,\lambda) = f(x,y)$$
So $F(x,y,\lambda)$ and $f(x,y)$ have the same 3-d surface, if you thought $a = F(x,y,\lambda)$ is a function in 3-d space. In fact the function $F(x,y,\lambda)$ is a function in 4-d space.
What we do is going to find the extremum of the 4-d space function $F(x,y,\lambda)$. When $F(x,y,\lambda)$ get its extremum, the conditions exactly fit the origin problem conditions, occasionally, luckily in fact.
Solving the simple 4-d problem solves the complex 3-d problem. “Why not solve the simple problem?”, Lagrange thought.

[MORE DETAILED EXPLANATION COMES SOON.]

reference

1 [youdaoDic]
特别的，我们学过,怎样找到直线与平面的交点,通过将参数方程带入平面方程。
And, well, we’ve learned in particular how to find where a line intersects a plane by plugging in the parametric equation into the equation of a plane.
2 拉格朗日乘数法_百度百科