Advanced Global Illumination Chapter3 —— Monte Carlo Methods

this chapter introduces the concept of Monte Carlo integration and reviews some basic concepts in probability theory. we also present techniques to create better distributions of samples.

3.1 brief history
the term “Monte Carlo” was coined in the 1940s, at the advant of electronic computing, to describe mathematical techniques that use statistical sampling to simulate phenomena or evaluate values of functions. these techniques were originally devised to simualte neutron transport by scientists such as Stanislaw Ulam, John von Neumann, and Nicholas Metropolis, among others, who were working on the development of nuclear 核 weapons. however, early examples of computations that can be defined as Monte Carlo exits, though without the use of computers to draw samples. one of the earliest documented examples of a Monte Carlo computation was done by Comte de Buffon in 1677. he conducted an experiment in which a needle of length L was thrown at random on a horionzontal plane with lines drawn at a distance d apart (d>L). 蒲丰投针试验 he repeated the experiment many times to estimate the the probability P that the needle would intersect one of these lines. he also analytically evaluated P as

laplace later sugguested that this technique of repeated experimentation could be used to compute an estimated vlaue of π. Kalos and Whitlock present early examples of Monte Carlo methods.

Monte Carlo 方法是计算机模拟的基础，它的名字来源于世界著名的赌城——摩纳哥的蒙特卡洛，其历史起源于1777年法国科学家蒲丰提出的一种计算圆周率的方法——随机投针法。

3.2 why are monte carlo techniques useful?
consider a problem that must be solved, for example, computing the value of the integration of a function with respect to an appropriately defined measure of domain. the monte carlo approach to solving this problem would be to define a random variable suh that the expected value of that random 期望值 variable would be the solution to the problem. samples of this random variable are then drawn and averaged to compute an estimate of the expected value of the random variable. this estimated expected value is an approximation of the solution of the problem we originally wanted to solve.

one major strength of the monte carlo approach lies in its conceptual simplicity; once an appropriate random variable is found, the computation consists of sampling the random variable and averaging the estimates obtained from the sample. another advantage of monte carlo tehniques is that they can be applied to a wide range of problems. it is intuitive that monte carlo techniques would apply to problems that are stochastic in nature, for example, transport problems in nuclear pysics. however, monte carlo techniques are applicable to an even wider range of problems, for example, problems that require the higher-dimensional integration of complicated functions. in fact, for these problems, monte carlo techniques are often the only feasible solution.

one disadvantages of monte carlo techniques is their relatively slow convergence rate of 1/sqr(N), where N is the number of samples (see section 3.4). as consequence, several variance reduction techniques have been developed in the field, discussed in this chapter. however, it should be noted that despite all these optimizations, monte carlo techniques still converge quite slowly and, therefore, are not used unless there are no viable alternatives. for example, even though monte carlo techniques are often illustrated using one-dimensional examples, they are not typially the most efficient solution techniuqes for problems of this kind. But there are problems for which Monte Carlo methods are the only feasible solution technique: higher-dimensional integrals and integrals with nonsmooth integrands, among others.

3.3 review of probability theory

in this section, we briefly review important concepts from probability theory. a monte carlo process is a sequence of random events. often, a numberical outcome can be associated with each possible event. for example, when a fair die is thrown, the outcome could be any value from 1 to 6. a random variable describes the possible outcomes of an experiment.

3.3.1 discrete random variables
when a random variable can take a finite number of possible values, it is called a discrete random variable. for a discrete random variable, a probability pi can be associated with any event with outcome xi.

a random variable xdie might be said to have a value of 1 to 6 associated with each of the possible outcomes of the throw of the die. the probability pi associated with each outcome for a fair die is 1/6.

some properties of the probabilities pi are:

1. The probablity of an event lies between 0 and 1: 0 ≤ pi ≤ 1. If an
   outcome never occurs, its probability is 0; if an event always occurs,
   its probability is 1.

2. The probability that either of two events occurs is:
   
   two events are mutually exclusive if and only if the occurence of one of the events implies the other event cannot possible occur. in the case of two such mutually exclusive events.
   

3. a set of all the possible events/outcomes of an experiment such that the events are mutually exclusive and collectively exhaustive satisfies the following normalization property:
   expected value
   for a discrete random variable with n possible outcomes, the expected value, or mean, of the random variable is
   

for the case of a fair die, the expected value of the die throws is