Laplace's method


In mathematics, Laplace's method, named after Pierre-Simon Laplace, is a technique used to approximate integrals of the form

where is a twice-differentiable function, M is a large number, and the endpoints a and b could possibly be infinite. This technique was originally presented in Laplace (1774).

In Bayesian statistics, Laplace approximation can refer to either approximating the posterior normalizing constant with Laplace's method[1] or approximating the posterior distribution with a Gaussian centered at the maximum a posteriori estimate.[2]

The idea of Laplace's methodEdit

  has a global maximum at 0.   is shown on top for M = 0.5, and at the bottom for M = 3 (both in blue). As M grows the approximation of this function by a Gaussian function (shown in red) improves. This observation underlies Laplace's method.

Suppose the function   has a unique global maximum at x0. Let   be a constant and consider the following two functions:


Note that x0 will be the global maximum of   and   as well. Now observe:


As M increases, the ratio for   will grow exponentially, while the ratio for   does not change. Thus, significant contributions to the integral of this function will come only from points x in a neighbourhood of x0, which can then be estimated.

General theory of Laplace's methodEdit

To state and motivate the method, we need several assumptions. We will assume that x0 is not an endpoint of the interval of integration, that the values   cannot be very close to   unless x is close to x0, and that  

We can expand   around x0 by Taylor's theorem,


where   (see: big O notation).

Since   has a global maximum at x0, and since x0 is not an endpoint, it is a stationary point, so the derivative of   vanishes at x0. Therefore, the function   may be approximated to quadratic order


for x close to x0 (recall  ). The assumptions ensure the accuracy of the approximation


(see the picture on the right). This latter integral is a Gaussian integral if the limits of integration go from −∞ to +∞ (which can be assumed because the exponential decays very fast away from x0), and thus it can be calculated. We find


A generalization of this method and extension to arbitrary precision is provided by Fog (2008).

Formal statement and proofEdit

Suppose   is a twice continuously differentiable function on   and there exists a unique point   such that:




Lower bound: Let  . Since   is continuous there exists   such that if   then   By Taylor's Theorem, for any  


Then we have the following lower bound:


where the last equality was obtained by a change of variables


Remember   so we can take the square root of its negation.

If we divide both sides of the above inequality by


and take the limit we get:


since this is true for arbitrary   we get the lower bound:


Note that this proof works also when   or   (or both).

Upper bound: The proof is similar to that of the lower bound but there are a few inconveniences. Again we start by picking an   but in order for the proof to work we need   small enough so that   Then, as above, by continuity of   and Taylor's Theorem we can find   so that if  , then


Lastly, by our assumptions (assuming   are finite) there exists an   such that if  , then  .

Then we can calculate the following upper bound:


If we divide both sides of the above inequality by


and take the limit we get:


Since   is arbitrary we get the upper bound:


And combining this with the lower bound gives the result.

Note that the above proof obviously fails when   or   (or both). To deal with these cases, we need some extra assumptions. A sufficient (not necessary) assumption is that for  


and that the number   as above exists (note that this must be an assumption in the case when the interval   is infinite). The proof proceeds otherwise as above, but with a slightly different approximation of integrals:


When we divide by


we get for this term


whose limit as   is  . The rest of the proof (the analysis of the interesting term) proceeds as above.

The given condition in the infinite interval case is, as said above, sufficient but not necessary. However, the condition is fulfilled in many, if not in most, applications: the condition simply says that the integral we are studying must be well-defined (not infinite) and that the maximum of the function at   must be a "true" maximum (the number   must exist). There is no need to demand that the integral is finite for   but it is enough to demand that the integral is finite for some  

This method relies on 4 basic concepts such as

1. Relative error

The “approximation” in this method is related to the relative error and not the absolute error. Therefore, if we set


the integral can be written as


where   is a small number when   is a large number obviously and the relative error will be


Now, let us separate this integral into two parts:   region and the rest.

2.   around the stationary point when   is large enough

Let’s look at the Taylor expansion of   around x0 and translate x to y because we do the comparison in y-space, we will get


Note that   because   is a stationary point. From this equation you will find that the terms higher than second derivative in this Taylor expansion is suppressed as the order of   so that   will get closer to the Gaussian function as shown in figure. Besides,

The figure of   with   equals 1, 2 and 3, and the red line is the curve of function   .
3. The larger   is, the smaller range of   is related

Because we do the comparison in y-space,   is fixed in   which will cause  ; however,   is inversely proportional to  , the chosen region of   will be smaller when   is increased.

4. If the integral in Laplace’s method converges, the contribution of the region which is not around the stationary point of the integration of its relative error will tend to zero as   grows.

Relying on the 3rd concept, even if we choose a very large Dy, sDy will finally be a very small number when   is increased to a huge number. Then, how can we guarantee the integral of the rest will tend to 0 when   is large enough?

The basic idea is to find a function   such that   and the integral of   will tend to zero when   grows. Because the exponential function of   will be always larger than zero as long as   is a real number, and this exponential function is proportional to   the integral of   will tend to zero. For simplicity, choose   as a tangent through the point   as shown in the figure:

  is denoted by the two tangent lines passing through  . When   gets smaller, the cover region will be larger.

If the interval of the integration of this method is finite, we will find that no matter   is continue in the rest region, it will be always smaller than   shown above when   is large enough. By the way, it will be proved later that the integral of   will tend to zero when   is large enough.

If the interval of the integration of this method is infinite,   and   might always cross to each other. If so, we cannot guarantee that the integral of   will tend to zero finally. For example, in the case of     will always diverge. Therefore, we need to require that   can converge for the infinite interval case. If so, this integral will tend to zero when   is large enough and we can choose this   as the cross of   and  

You might ask that why not choose   as a convergent integral? Let me use an example to show you the reason. Suppose the rest part of   is   then   and its integral will diverge; however, when   the integral of   converges. So, the integral of some functions will diverge when   is not a large number, but they will converge when   is large enough.

Based on these four concepts, we can derive the relative error of this Laplace's method.

Other formulationsEdit

Laplace's approximation is sometimes written as


where   is positive.

Importantly, the accuracy of the approximation depends on the variable of integration, that is, on what stays in   and what goes into  .[3]

The derivation of its relative error

First, use   to denote the global maximum, which will simplify this derivation. We are interested in the relative error, written as  ,




So, if we let


and  , we can get


since  .

For the upper bound, note that   thus we can separate this integration into 5 parts with 3 different types (a), (b) and (c), respectively. Therefore,


where   and   are similar, let us just calculate   and   and   are similar, too, I’ll just calculate  .

For  , after the translation of  , we can get


This means that as long as   is large enough, it will tend to zero.

For  , we can get




and   should have the same sign of   during this region. Let us choose   as the tangent across the point at   , i.e.   which is shown in the figure

  is the tangent lines across the point at   .

From this figure you can find that when   or   gets smaller, the region satisfies the above inequality will get larger. Therefore, if we want to find a suitable   to cover the whole   during the interval of  ,   will have an upper limit. Besides, because the integration of   is simple, let me use it to estimate the relative error contributed by this  .

Based on Taylor expansion, we can get




and then substitute them back into the calculation of  ; however, you can find that the remainders of these two expansions are both inversely proportional to the square root of  , let me drop them out to beautify the calculation. Keeping them is better, but it will make the formula uglier.


Therefore, it will tend to zero when   gets larger, but don't forget that the upper bound of   should be considered during this calculation.

About the integration near  , we can also use Taylor's Theorem to calculate it. When  


and you can find that it is inversely proportional to the square root of  . In fact,   will have the same behave when   is a constant.

Conclusively, the integral near the stationary point will get smaller as   gets larger, and the rest parts will tend to zero as long as   is large enough; however, we need to remember that   has an upper limit which is decided by whether the function   is always larger than   in the rest region. However, as long as we can find one   satisfying this condition, the upper bound of   can be chosen as directly proportional to   since   is a tangent across the point of   at  . So, the bigger   is, the bigger   can be.

In the multivariate case where   is a  -dimensional vector and   is a scalar function of  , Laplace's approximation is usually written as:


where   is the Hessian matrix of   evaluated at   and where   denotes matrix determinant. Analogously to the univariate case, the Hessian is required to be negative definite.[4]

By the way, although   denotes a  -dimensional vector, the term   denotes an infinitesimal volume here, i.e.  .

Laplace's method extension: Steepest descentEdit

In extensions of Laplace's method, complex analysis, and in particular Cauchy's integral formula, is used to find a contour of steepest descent for an (asymptotically with large M) equivalent integral, expressed as a line integral. In particular, if no point x0 where the derivative of   vanishes exists on the real line, it may be necessary to deform the integration contour to an optimal one, where the above analysis will be possible. Again the main idea is to reduce, at least asymptotically, the calculation of the given integral to that of a simpler integral that can be explicitly evaluated. See the book of Erdelyi (1956) for a simple discussion (where the method is termed steepest descents).

The appropriate formulation for the complex z-plane is


for a path passing through the saddle point at z0. Note the explicit appearance of a minus sign to indicate the direction of the second derivative: one must not take the modulus. Also note that if the integrand is meromorphic, one may have to add residues corresponding to poles traversed while deforming the contour (see for example section 3 of Okounkov's paper Symmetric functions and random partitions).

Further generalizationsEdit

An extension of the steepest descent method is the so-called nonlinear stationary phase/steepest descent method. Here, instead of integrals, one needs to evaluate asymptotically solutions of Riemann–Hilbert factorization problems.

Given a contour C in the complex sphere, a function   defined on that contour and a special point, say infinity, one seeks a function M holomorphic away from the contour C, with prescribed jump across C, and with a given normalization at infinity. If   and hence M are matrices rather than scalars this is a problem that in general does not admit an explicit solution.

An asymptotic evaluation is then possible along the lines of the linear stationary phase/steepest descent method. The idea is to reduce asymptotically the solution of the given Riemann–Hilbert problem to that of a simpler, explicitly solvable, Riemann–Hilbert problem. Cauchy's theorem is used to justify deformations of the jump contour.

The nonlinear stationary phase was introduced by Deift and Zhou in 1993, based on earlier work of Its. A (properly speaking) nonlinear steepest descent method was introduced by Kamvissis, K. McLaughlin and P. Miller in 2003, based on previous work of Lax, Levermore, Deift, Venakides and Zhou. As in the linear case, "steepest descent contours" solve a min-max problem. In the nonlinear case they turn out to be "S-curves" (defined in a different context back in the 80s by Stahl, Gonchar and Rakhmanov).

The nonlinear stationary phase/steepest descent method has applications to the theory of soliton equations and integrable models, random matrices and combinatorics.

Laplace's method generalization: Median-point approximationEdit

In the generalization, evaluation of the integral is considered equivalent to finding the norm of the distribution with density


Denoting the cumulative distribution  , if there is a diffeomorphic Gaussian distribution with density


the norm is given by


and the corresponding diffeomorphism is


where   denotes cumulative standard normal distribution function.

In general, any distribution diffeomorphic to the Gaussian distribution has density


and the median-point is mapped to the median of the Gaussian distribution. Matching the logarithm of the density functions and their derivatives at the median point up to a given order yields a system of equations that determine the approximate values of   and  .

The approximation was introduced in 2019 by D. Makogon and C. Morais Smith primarily in the context of partition function evaluation for a system of interacting fermions.

Complex integralsEdit

For complex integrals in the form:


with   we make the substitution t = iu and the change of variable   to get the bilateral Laplace transform:


We then split g(c + ix) in its real and complex part, after which we recover u = t/i. This is useful for inverse Laplace transforms, the Perron formula and complex integration.

Example: Stirling's approximationEdit

Laplace's method can be used to derive Stirling's approximation


for a large integer N.

From the definition of the Gamma function, we have


Now we change variables, letting   so that   Plug these values back in to obtain


This integral has the form necessary for Laplace's method with


which is twice-differentiable:


The maximum of   lies at z0 = 1, and the second derivative of   has the value −1 at this point. Therefore, we obtain


See alsoEdit


  1. ^ Tierney, Luke; Kadane, Joseph B. (1986). "Accurate Approximations for Posterior Moments and Marginal Densities". J. Amer. Stat. Assoc. 81 (393): 82–86.
  2. ^ Gelman, Andrew; et al. (2014). "Modal and Distributional Approximations". Bayesian Data Analysis (Third ed.). Taylor & Francis. pp. 311–350. ISBN 978-1-4398-4095-5.
  3. ^ Butler, Ronald W (2007). Saddlepoint approximations and applications. Cambridge University Press. ISBN 978-0-521-87250-8.
  4. ^ MacKay, David J. C. (September 2003). Information Theory, Inference and Learning Algorithms. Cambridge: Cambridge University Press. ISBN 9780521642989.


  • Azevedo-Filho, A.; Shachter, R. (1994), "Laplace's Method Approximations for Probabilistic Inference in Belief Networks with Continuous Variables", in Mantaras, R.; Poole, D. (eds.), Uncertainty in Artificial Intelligence, San Francisco, CA: Morgan Kaufmann, CiteSeerX
  • Deift, P.; Zhou, X. (1993), "A steepest descent method for oscillatory Riemann–Hilbert problems. Asymptotics for the MKdV equation", Ann. of Math., vol. 137, no. 2, pp. 295–368, arXiv:math/9201261, doi:10.2307/2946540, JSTOR 2946540.
  • Erdelyi, A. (1956), Asymptotic Expansions, Dover.
  • Fog, A. (2008), "Calculation Methods for Wallenius' Noncentral Hypergeometric Distribution", Communications in Statistics, Simulation and Computation, vol. 37, no. 2, pp. 258–273, doi:10.1080/03610910701790269.
  • Laplace, P S (1774), "Mémoires de Mathématique et de Physique, Tome Sixième" [Memoir on the probability of causes of events.], Statistical Science, 1 (3): 366–367, JSTOR 2245476
  • Wang, Xiang-Sheng; Wong, Roderick (2007). "Discrete analogues of Laplace's approximation". Asymptot. Anal. 54 (3–4): 165–180.

This article incorporates material from saddle point approximation on PlanetMath, which is licensed under the Creative Commons Attribution/Share-Alike License.