Stochastic Process -

Japanese: 確率過程 - かくりつかてい
Stochastic Process -

The yen exchange rate, the amount of rainfall in a certain area, noise phenomena in communications engineering, Brownian motion, etc. are phenomena that change with chance over time. A mathematical model of such phenomena is a stochastic process. Mathematically speaking, a family of random variables {X t } with real number t as a parameter is called a stochastic process. Here, if t is fixed, X t is a random variable, but the sample space Ω of X t is constant and unrelated to t. Since X t is a function of t and ω, an element of Ω, X t = X(t,ω)
When the element ω of Ω is fixed, X(t,ω) becomes a function of t, and this function is called a sample function or path.

Arbitrarily select a finite number of real numbers t1 , t2 , ..., tn and consider a k-dimensional random variable ( Xt1 , Xt2 , ..., Xtk ). The probability distribution of this k-dimensional random variable is called the finite-dimensional distribution of the stochastic process { Xt }. Stochastic processes can be divided into additive processes, Markov processes, stationary processes, etc., depending on the properties of their finite-dimensional distribution. For the interval I = (a, b), the random variable XI is XI = Xb - Xa
We define the intervals I,J,……,K as follows:
For random variables X I ,X J ,……,X K
When are independent, the stochastic process {X t } is called an additive process.

In an additive process, when the probability of all ω where the sample function is a continuous function of t is 1, the probability distribution of the random variable X t -X s (t>s) is normal. This type of stochastic process is called a normal additive process. In particular, when the mean of X t -X s is 0 and the variance is t-s, this stochastic process is called a Wiener process or Wiener's Brownian motion. Also, when the probability of all ω where the sample function is a step function that increases in jumps of height 1 is 1, the probability distribution of the random variable X t -X s is Poisson distribution. This type of stochastic process is called a Poisson process.

Next, in a time-changing random phenomenon, when the state at time s is known, the state at time t (t>s) may be determined only by the state at time s, and may be unrelated to the state prior to s. A stochastic process with this property is called a Markov process. In more detail, if {X t } is a stochastic process, then
s 1 <s 2 <...<s n <t
If the conditional probability law for Xt given the values ​​of Xs1 , Xs2 , …, Xsn , and Xt is equal to the conditional probability law for Xt given only the value of Xsn , that is, P( Xt (ω)∈A| Xs1 = xs1 , …, Xsn = xsn )
=P(X t (ω)∈A|X sn =x sn )
When this holds, {X t } is called a Markov process.

The probability law of a Markov process is determined by the conditional probability F(s,x;t,A) of Xt when Xs = x, with s≦t. This probability is called the transition probability.

[Shigeru Furuya]

"Modern Mathematics 14: Probability Theory" by Kiyoshi Ito (1953, Iwanami Shoten)

[Reference] | Random variables | Stationary processes

Source: Shogakukan Encyclopedia Nipponica About Encyclopedia Nipponica Information | Legend

Japanese:

円の交換レート、ある地区における降雨量、通信工学における雑音現象、ブラウン運動などは、時の経過に伴って偶然性をもって変動する現象である。このような現象の数学的モデルが確率過程である。数学的にいえば、実数tをパラメーターとする確率変数族{Xt}を確率過程という。ここでtを固定すればXtは確率変数であるが、Xtの標本空間Ωはtに無関係で一定のものである。XtはtとΩの元ωの関数であるから
  Xt=X(t,ω)
と表される。Ωの元ωを固定するとX(t,ω)はtの関数となるが、この関数を標本関数または道とよぶ。

 任意に有限個の実数t1、t2、……、tnを選んで、k次元確率変数(Xt1,Xt2,……,Xtk)を考え、このk次元確率変数の確率分布を確率過程{Xt}の有限次元分布という。確率過程はその有限次元分布の性質によって、加法過程、マルコフ過程、定常過程などに分けられる。区間I=(a,b)に対して確率変数XI
  XI=Xb-Xa
と定める。互いに重なり合うことのない区間
  I,J,……,K
に対して、確率変数
  XI,XJ,……,XK
が独立であるとき、確率過程{Xt}を加法過程という。

 加法過程において、標本関数がtの連続関数であるようなω全体の確率が1であるとき、確率変数Xt-Xs(t>s)の確率分布は正規分布となる。このような確率過程を正規加法過程という。ここでとくにXt-Xsの平均値が0で、分散がt-sであるとき、この確率過程をウィーナー過程またはウィーナーのブラウン運動という。また、標本関数が高さ1の飛躍で増加するような階段関数であるようなω全体の確率が1であるとき、確率変数Xt-Xsの確率分布はポアソン分布である。このような確率過程をポアソン過程という。

 次に、時間とともに変化する偶然的現象において、時刻sにおける状態がわかっているとき、時刻t(t>s)における状態が、時刻sにおける状態だけで決まり、s以前の状態には無関係なことがある。このような性質をもつ確率過程をマルコフ過程という。詳しくいえば、{Xt}を確率過程とするとき、
  s1<s2<……<sn<t
としてXs1、Xs2、……、Xsn、Xtの値を与えたときのXtの条件つき確率法則が、Xsnの値だけを与えたときのXtの条件つき確率法則に等しいとき、すなわち
  P(Xt(ω)∈A|Xs1=xs1,……,Xsn=xsn)
  =P(Xt(ω)∈A|Xsn=xsn)
が成り立つとき{Xt}をマルコフ過程という。

 マルコフ過程の確率法則はs≦tとしてXs=xのときのXtの条件つき確率F(s,x;t,A)によって決定される。この確率を遷移確率という。

[古屋 茂]

『伊藤清著『現代数学14 確率論』(1953・岩波書店)』

[参照項目] | 確率変数 | 定常過程

出典 小学館 日本大百科全書(ニッポニカ)日本大百科全書(ニッポニカ)について 情報 | 凡例

<<:  Stochastic control

>>:  Probability - kakuritsu (English spelling) probability

Recommend

Shu

[Sound] Shu (Shifu) (Han) Ju (Jifu) (Wu) [Reading]...

van de Cappelle, J.

...Furthermore, the cutthroat free competition to...

Doodle

...After the 1973 American film American Graffiti...

Mr. Samejima - Mr. Samejima

A medieval samurai family in Suruga and Satsuma. A...

Nishiki [town] - Nishiki

A former town in Taki District, located in the nor...

Ephippianthus sawadanus (F.Maek.) Ohwi

A small terrestrial orchid found rarely on the for...

Experiments in Art and Technology

…A group of artists and engineers formed in New Y...

cheliped

…The terminal segment is a movable finger that in...

harpoon

...The iron harpoon heads used for this are basic...

Yoshishi

…One of the ancient Japanese surnames. Also writt...

《Sweeping Away》 - Even More

…This culture fostered the creation of the 13 Hok...

Circumcision parent - Katsurei oya

...However, female circumcision is common among t...

Phrase - Hyosoku

In classical Chinese poetry, this is a set of rul...

Reforma Oil Field - Reforma (English)

In 1957, an oil field was discovered in the vicini...

Alexandrov, GV - Alexandrov

...Silent films such as Eisenstein's Battlesh...