Lagrange multipliers

From Wikipedia, the free encyclopedia

Fig. 1. Drawn in red is the locus (contour) of points satisfying the constraint g(x,y) = c. Drawn in blue are contours of f. Arrows represent the gradient, which points in a direction normal to the contour.

Fig. 2. Same problem as above as three-dimensional visualization, assuming that we want to maximize

f (x, y)

In mathematical optimization problems, the method of Lagrange multipliers, named after Joseph Louis Lagrange, is a method for finding the extrema of a function of several variables subject to one or more constraints; it is the basic tool in nonlinear constrained optimization.

Simply put, the technique is able to determine where on a particular set of points (such as a circle, sphere, or plane) a particular function is the smallest (or largest).

More formally, Lagrange multipliers compute the stationary points of the constrained function. By Fermat's theorem, extrema occur either at these points, or on the boundary, or at points where the function is not differentiable.

It reduces finding stationary points of a constrained function in n variables with k constraints to finding stationary points of an unconstrained function in n+k variables. The method introduces a new unknown scalar variable (called the Lagrange multiplier) for each constraint, and defines a new function (called the Lagrangian) in terms of the original function, the constraints, and the Lagrange multipliers.

1 Introduction
- 1.1 Justification
- 1.2 Caveat: extrema versus stationary points
2 A more general formulation: The weak Lagrangian principle
3 Interpretation of λ_i
4 Examples
5 Economics
6 The strong Lagrangian principle: Lagrange duality
7 See also
8 References
9 External links

[] Introduction

Consider a two-dimensional case. Suppose we have a function $f (x, y)$ we wish to maximize or minimize subject to the constraint

$gleft( x, y right) = c,$

where c is a constant. We can visualize contours of $f$ given by

$f left( x, y right)=d_n$

for various values of $d n$ , and the contour of $g$ given by $g (x, y) = c$ .

Suppose we walk along the contour line with $g = c$ . In general the contour lines of $f$ and $g$ may be distinct, so traversing the contour line for $g = c$ could intersect with or cross the contour lines of $f$ . This is equivalent to saying that while moving along the contour line for $g = c$ the value of $f$ can vary. Only when the contour line for $g = c$ touches contour lines of $f$ tangentially, we do not increase or decrease the value of $f$ - that is, when the contour lines touch but do not cross.

This occurs exactly when the tangential component of the total derivative vanishes: $df_parallel = 0$ , which is at the constrained stationary points of $f$ (which include the constrained local extrema, assuming $f$ is differentiable). Computationally, this is when the gradient of $f$ is normal to the constraint(s): when $nabla f = lambda nabla g$ for some scalar $λ$ (where $nabla$ is the gradient). Note that the constant $λ$ is required because, even though the directions of both gradient vectors are equal, the magnitudes of the gradient vectors are most likely not equal.

A familiar example can be obtained from weather maps, with their contour lines for temperature and pressure: the constrained extrema will occur where the superposed maps show touching lines (isopleths).

Geometrically we translate the tangency condition to saying that the gradients of $f$ and $g$ are parallel vectors at the maximum, since the gradients are always normal to the contour lines. Thus we want points $(x, y)$ where $g (x, y) = c$ and

$nabla_{x,y} f = lambda nabla_{x,y} g$ ,

where

$nabla_{x,y} = left( frac{partial}{partial x}, frac{partial}{partial y} right).$

To incorporate these conditions into one equation, we introduce an auxiliary function

$F left( x , y, lambda right) = f left(x, y right) + lambda left(g left(x, y right) - c right),$

and solve

$nabla_{x,y,lambda} F left( x , y, lambda right)=0$ .

[] Justification

As discussed above, we are looking for stationary points of $f$ seen while travelling on the level set $g (x, y) = c$ . This occurs just when the gradient of $f$ has no component tangential to the level sets of $g$ . This condition is equivalent to $nabla_{x,y} f(x,y) = lambda nabla_{x,y} g(x,y)$ for some $λ$ . Stationary points $(x, y,λ)$ of $F$ also satisfy $g (x, y) = c$ as can be seen by considering the derivative with respect to $λ$ .

[] Caveat: extrema versus stationary points

Be aware that the solutions are the stationary points of the Lagrangian $F$ , and are saddle points: they are not necessarily extrema of $F$ . $F$ is unbounded: given a point $(x, y)$ that doesn't lie on the constraint, letting $lambda to pm infty$ makes $F$ arbitrarily large or small. However, under certain stronger assumptions, as we shall see below, the strong Lagrangian principle holds, which states that the maxima of $f$ maximize the Lagrangian globally.

[] A more general formulation: The weak Lagrangian principle

Denote the objective function by $f(mathbf x)$ and let the constraints be given by $g_k(mathbf x)=0$ , perhaps by moving constants to the left, as in $h_k(mathbf x)-c_k=g_k(mathbf x)$ . The domain of f should be an open set containing all points satisfying the constraints. Furthermore, $f$ and the $g k$ must have continuous first partial derivatives and the gradients of the $g k$ must not be zero on the domain.^[1] Now, define the Lagrangian, $Λ$ , as

$Lambda(mathbf x, boldsymbol lambda) = f + sum_k lambda_k g_k.$

k

is an index for variables and functions associated with a particular constraint,

k

$boldsymbollambda$ without a subscript indicates the vector with elements $mathbf lambda_k$ , which are taken to be independent variables.

Observe that both the optimization criteria and constraints $g k (x)$ are compactly encoded as stationary points of the Lagrangian:

$nabla_{mathbf x} Lambda = mathbf{0}$ if and only if $nabla_{mathbf x} f = - sum_k lambda_k nabla_{mathbf x} g_k,$

$nabla_{mathbf x}$ means to take the gradient only with respect to each element in the vector $mathbf x$ , instead of all variables.

and

$nabla_{mathbf lambda} Lambda = mathbf{0}$ implies

g k = 0.

Collectively, the stationary points of the Lagrangian,

$nabla Lambda = mathbf{0}$ ,

give a number of unique equations totaling the length of $mathbf x$ plus the length of $mathbf lambda$ .

[] Interpretation of $λ i$

Often the Lagrange multipliers have an interpretation as some salient quantity of interest. To see why this might be the case, observe that:

$frac{partial Lambda}{partial {g_k}} = lambda_k.$

So, λ_k is the rate of change of the quantity being optimized as a function of the constraint variable. As examples, in Lagrangian mechanics the equations of motion are derived by finding stationary points of the action, the time integral of the difference between kinetic and potential energy. Thus, the force on a particle due to a scalar potential, $F=-nabla V$ , can be interpreted as a Lagrange multiplier determining the change in action (transfer of potential to kinetic energy) following a variation in the particle's constrained trajectory. In economics, the optimal profit to a player is calculated subject to a constrained space of actions, where a Lagrange multiplier is the value of relaxing a given constraint (e.g. through bribery or other means).

The method of Lagrange multipliers is generalized by the Karush-Kuhn-Tucker conditions.

[] Examples

[] Very simple example

Fig. 3. Illustration of the constrained optimization problem.

Suppose you wish to maximize $f (x, y) = x + y$ subject to the constraint $x 2 + y 2 = 1$ . The constraint is the unit circle, and the level sets of f are diagonal lines (with slope -1), so one can see graphically that the maximum occurs at $(sqrt{2}/2,sqrt{2}/2)$ (and the minimum occurs at $(-sqrt{2}/2,-sqrt{2}/2)$

Formally, set $g (x, y) = x 2 + y 2 - 1$ , and

Λ(x, y,λ) = f (x, y) + λ g (x, y) = x + y + λ(x 2 + y 2 - 1)

Set the derivative $d Λ = 0$ , which yields the system of equations:

$begin{align} frac{partial Lambda}{partial x} &= 1 + 2 lambda x &&= 0, qquad text{(i)} \ frac{partial Lambda}{partial y} &= 1 + 2 lambda y &&= 0, qquad text{(ii)} \ frac{partial Lambda}{partial lambda} &= x^2 + y^2 - 1 &&= 0, qquad text{(iii)} end{align}$

As always, the $partial lambda$ equation is the original constraint.

Combining the first two equations yields $x = y$ (explicitly, $x neq 0$ ,otherwise (i) yields 1 = 0), so one can solve for $λ$ , yielding $λ = - 1 / (2 x)$ , which one can substitute into (ii)).

Substituting into (iii) yields $2 x 2 = 1$ , so $x=pm sqrt{2}/2$ and the stationary points are $(sqrt{2}/2,sqrt{2}/2)$ and $(-sqrt{2}/2,-sqrt{2}/2)$ . Evaluating the objective function f on these yields

$f(sqrt{2}/2,sqrt{2}/2)=sqrt{2}mbox{ and } f(-sqrt{2}/2, -sqrt{2}/2)=-sqrt{2},$

thus the maximum is $sqrt{2}$ , which is attained at $(sqrt{2}/2,sqrt{2}/2)$ and the minimum is $-sqrt{2}$ , which is attained at $(-sqrt{2}/2,-sqrt{2}/2)$ .

[] Simple example

Fig. 4. Illustration of the constrained optimization problem.

Suppose you want to find the maximum values for

$f(x, y) = x^2 y ,$

with the condition that the x and y coordinates lie on the circle around the origin with radius √3, that is,

$x^2 + y^2 = 3. ,$

As there is just a single condition, we will use only one multiplier, say λ.

Use the constraint to define a function g(x, y):

$g (x, y) = x^2 +y^2 -3. ,$

The function g is identically zero on the circle of radius √3. So any multiple of g(x, y) may be added to f(x, y) leaving f(x, y) unchanged in the region of interest (above the circle where our original constraint is satisfied). Let

$Lambda(x, y, lambda) = f(x,y) + lambda g(x, y) = x^2y + lambda (x^2 + y^2 - 3). ,$

The critical values of $Λ$ occur when its gradient is zero. The partial derivatives are

$begin{align} frac{partial Lambda}{partial x} &= 2 x y + 2 lambda x &&= 0, qquad text{(i)} \ frac{partial Lambda}{partial y} &= x^2 + 2 lambda y &&= 0, qquad text{(ii)} \ frac{partial Lambda}{partial lambda} &= x^2 + y^2 - 3 &&= 0. qquad text{(iii)} end{align}$

Equation (iii) is just the original constraint. Equation (i) implies $x = 0$ or λ = −y. In the first case, if $x = 0$ then we must have $y = pm sqrt{3}$ by (iii) and then by (ii) λ=0. In the second case, if λ = −y and substituting into equation (ii) we have that,

$x^2 - 2y^2 = 0. ,$

Then x² = 2y². Substituting into equation (iii) and solving for y gives this value of y:

$y = pm 1. ,$

Thus there are six critical points:

$(sqrt{2},1); quad (-sqrt{2},1); quad (sqrt{2},-1); quad (-sqrt{2},-1); quad (0,sqrt{3}); quad (0,-sqrt{3}).$

Evaluating the objective at these points, we find

$f(pmsqrt{2},1) = 2; quad f(pmsqrt{2},-1) = -2; quad f(0,pm sqrt{3})=0.$

Therefore, the objective function attains a global maximum (with respect to the constraints) at $(pmsqrt{2},1)$ and a global minimum at $(pmsqrt{2},-1).$ The point $(0,sqrt{3})$ is a local minimum and $(0,-sqrt{3})$ is a local maximum.

[] Example: entropy

Suppose we wish to find the discrete probability distribution with maximal information entropy. Then

$f(p_1,p_2,ldots,p_n) = -sum_{k=1}^n p_klog_2 p_k.$

Of course, the sum of these probabilities equals 1, so our constraint is g(p) = 1 with

$g(p_1,p_2,ldots,p_n)=sum_{k=1}^n p_k.$

We can use Lagrange multipliers to find the point of maximum entropy (depending on the probabilities). For all k from 1 to n, we require that

$frac{partial}{partial p_k}(f+lambda (g-1))=0,$

which gives

$frac{partial}{partial p_k}left(-sum_{k=1}^n p_k log_2 p_k + lambda (sum_{k=1}^n p_k - 1) right) = 0.$

Carrying out the differentiation of these n equations, we get

$-left(frac{1}{ln 2}+log_2 p_k right) + lambda = 0.$

This shows that all p_i are equal (because they depend on λ only). By using the constraint ∑_k p_k = 1, we find

$p_k = frac{1}{n}.$

Hence, the uniform distribution is the distribution with the greatest entropy.

[] Economics

Constrained optimization plays a central role in economics. For example, the choice problem for a consumer is represented as one of maximizing a utility function subject to a budget constraint. The Lagrange multiplier has an economic interpretation as the shadow price associated with the constraint, in this case the marginal utility of income.

[] The strong Lagrangian principle: Lagrange duality

Given a convex optimization problem in standard form

$begin{align} text{minimize } &f_0(x) \ text{subject to } &f_i(x) leq 0, i in left {1,dots,m right } \ &h_i(x) = 0, i in left {1,dots,p right } end{align}$

with the domain $mathcal{D} subset mathbb{R}^n$ having non-empty interior, the Lagrangian function $L: mathbb{R}^n times mathbb{R}^m times mathbb{R}^p to mathbb{R}$ is defined as

$L(x,lambda,nu) = f_0(x) + sum_{i=1}^m lambda_i f_i(x) + sum_{i=1}^p nu_i h_i(x).$

The vectors $λ$ and $ν$ are called the dual variables or Lagrange multiplier vectors associated with the problem. The Lagrange dual function $g:mathbb{R}^m times mathbb{R}^p to mathbb{R}$ is defined as

$g(lambda,nu) = inf_{xinmathcal{D}} L(x,lambda,nu) = inf_{xinmathcal{D}} left ( f_0(x) + sum_{i=1}^m lambda_i f_i(x) + sum_{i=1}^p nu_i h_i(x) right ).$

The dual function $g$ is concave, even when the initial problem is not convex. The dual function yields lower bounds on the optimal value $p *$ of the initial problem; for any $lambda geq 0$ and any $ν$ we have $g(lambda,nu) leq p^*$ . If a constraint qualification such as Slater's condition holds and the original problem is convex, then we have strong duality, i.e. $d^* = max_{lambda ge 0, nu} g(lambda,nu) = inf f_0 = p^*$ .

[] See also

Karush-Kuhn-Tucker conditions: generalization of the method of Lagrange multipliers.
Lagrange multipliers on Banach spaces: another generalization of the method of Lagrange multipliers.