Monte Carlo Method

(redirected from Monte Carlo analysis)
Also found in: Dictionary, Medical, Financial, Acronyms.
Related to Monte Carlo analysis: Sensitivity analysis

Monte Carlo method

[′män·tē ′kär·lō ‚meth·əd]
(statistics)
A technique which obtains a probabilistic approximation to the solution of a problem by using statistical sampling techniques.

Monte Carlo Method

 

a numerical method for solving mathematical problems using the modeling of random processes and events. The term “Monte Carlo method” was coined in 1949, although some calculations using the modeling of random events had been previously performed by statisticians. (The method is named after the city of Monte Carlo, which is well known for its Casino.) The Monte Carlo method gained wide-spread use only after the development of high-speed computers. Programs for carrying out calculations on computers using the Monte Carlo method are comparatively simple and, as a rule, do not require large-capacity internal memory.


Monte Carlo Method

 

a method in computational and applied mathematics based on the simulation of random variables and the construction of statistical estimates of desired values.

The Monte Carlo method is generally considered to have been formulated in 1944 by the American scientists J. von Neumann and S. Ulam. In connection with work on the development of atomic reactors, von Neumann and Ulam made extensive use of the apparatus of probability theory in the computer solution of applied problems. At first the Monte Carlo method was used primarily to solve complex problems of radiation transport theory and neutron physics for which traditional numerical methods were not suited. The method was subsequently applied to a large number of different problems in statistical mechanics. The method is now used in such areas as game theory, queuing theory, mathematical economics, and the theory of message transmission in the presence of interference.

To solve a deterministic problem by the Monte Carlo method, a probabilistic model is constructed, and the desired quantity, for example, a multidimensional integral, is represented as the mathematical expectation of a function of a stochastic process. The process is then simulated on a computer. Probabilistic models are known, for example, for the computation of integrals, the solving of integral equations of the second kind, the solving of systems of linear algebraic equations, the solving of boundary problems for elliptic equations, and the estimation of the eigenvalues of linear operators. An estimate with a low degree of error can be obtained by proper selection of the probabilistic model.

The simulation of random variables with given distributions plays a special role in various applications of the Monte Carlo method. Such simulation is generally done by transforming one or more independent values of a random number a that is uniformly distributed in the interval (0, 1). The sequences of sample values of a are usually obtained with a computer through the use of theoretical numerical algorithms, among which the residue method has become the most common. Such numbers are said to be pseudorandom; they are checked through statistical tests and the solution of standard problems.

If, in an estimate based on the Monte Carlo method, the random variables involved are determined by the actual process being simulated, then we speak of direct modeling. Such an estimate is inefficient if rare events are involved because the actual process contains little information about such events. This inefficiency is usually manifested in too large a probabilistic error (variance) of the random estimates of the desired quantities. Many techniques have been developed for reducing the variance of these estimates within the framework of the Monte Carlo method. Almost all the techniques are based on modifications of the simulation through the use of information on the function of the random variables whose expectation is being computed.

The Monte Carlo method has had, and continues to have, a considerable influence on the development of other methods of computational mathematics—for example, methods of numerical integration. The Monte Carlo method has been successfully used as a supplementary method in combination with other methods to solve many problems. (See also.)

REFERENCES

Metod Monte-Karlo v problème perenosa izluchenii. Moscow, 1967.
Metodstatisticheskikh ispytanii (Metod Monte-Karlo). Moscow, 1962.
Reshenie priamykh i nekotorykh obratnykh zadach atmosfernoi optiki metodom Monte-Karlo. Novosibirsk, 1968.
Ermakov, S. M. Metod Monte-Karlo i smezhnye voprosy. Moscow, 1971.
Mikhailov, G. A. Nekotorye voprosy teorii metodov Monte-Karlo. Novosibirsk, 1974.

G. I. MARCHUK

Monte Carlo method

A technique that provides approximate solutions to problems expressed mathematically. Using random numbers and trial and error, it repeatedly calculates the equations to arrive at a solution. Many of the Monte Carlo methods and practices used to be referred to as rather generic "statistical sampling." Monte Carlo, of course, is a historical reference to the famous casino in Monaco.
References in periodicals archive ?
Table 1 Monte Carlo Analysis Assumptions Most Maximum Minimum Likely Total price today $6,000 $4,000 $5,000 Cash rent today $170 $150 $160 Risky real rate 9% 7% 8% Development price $30,000 $15,000 $20,000 Table 2 Time to Development--Monte Carlo Analysis Results Years to Development Mean 24.
Our new tool is also unique in that its Monte Carlo analysis runs 5,000 simulations covering 15 different asset classes," she added.
Our solution complements traditional Monte Carlo analysis, which is the only analysis that's been available to analog and custom IC designers for years," said Mr.
In four minutes, CARMA was able to evaluate the current and future exposure to a counterparty with 230 open positions, using 2,500-path Monte Carlo analysis over an 8-year time horizon -- evidence of the practicality of near real-time credit risk analysis," said David Gilbert.
0's unique autostack VSR (Virtual Stack Realisation) engine performs a Monte Carlo analysis on 10,000 virtual builds of each stack to provide a statistical profile of board thickness and an accurate prediction of production yields.
The traditional way to verify the robustness of A/MS designs over process and environment variations is to run a Monte Carlo analysis on transistor-level descriptions.
Using both an historical analysis and a Monte Carlo analysis, their charts illustrate that for all time periods and for all investment portfolios in the study, the addition of the fixed annuity leads to better results.
T-Spice offers options and commands not found in Berkeley SPICE or most derivatives, such as design optimization, Monte Carlo analysis, multi-dimensional parameters, or source and temperature sweeping.