{
  "id": "kolmogorov-forward-and-backward-equations",
  "title": "Kolmogorov Forward and Backward Equations",
  "lang": "en",
  "source_lang": "en",
  "summary": "Kolmogorov forward and backward equations describe how Markov transition probabilities, probability distributions, and conditional expectations evolve in time.",
  "category": "mathematics",
  "subcategory": "probability-statistics",
  "type": "concept",
  "aliases": [
    "Kolmogorov equations",
    "forward and backward evolution equations"
  ],
  "generated_at": "2026-11-05",
  "model": "pro/gpt-6.1-sol",
  "url": "https://aiwiki.page/en/kolmogorov-forward-and-backward-equations/",
  "translations": {
    "zh-Hans": {
      "url": "https://aiwiki.page/zh/kolmogorov-forward-and-backward-equations/",
      "json": "https://aiwiki.page/zh/kolmogorov-forward-and-backward-equations.json"
    }
  },
  "sections": [
    {
      "heading": "",
      "text": "The Kolmogorov forward and backward equations are differential equations governing the transition probabilities of a stochastic process with the Markov property. They express the same probabilistic evolution from two complementary perspectives: the forward equation propagates a probability distribution toward later times, while the backward equation propagates a future observable or payoff toward earlier starting times. Both are determined by an infinitesimal operator that describes the process over very short time intervals.\n\nFor a continuous-time Markov chain, the equations are systems of ordinary differential equations involving transition rates. For a diffusion, they become partial differential equations involving drift and local covariance. The forward diffusion equation is commonly called the Fokker–Planck equation; its backward counterpart is central to exit probabilities, expected passage times, stochastic control, and probabilistic representations of differential equations. Their common structure connects probability theory with operator theory, statistical physics, population modeling, and numerical analysis."
    },
    {
      "heading": "Transition probabilities and infinitesimal generators",
      "text": "Let \\(X_t\\) be a Markov process on a state space \\(E\\). Its transition kernel is\n\n\\[P_{s,t}(x,A)=\\Pr(X_t\\in A\\mid X_s=x), \\qquad s\\leq t,\\]\n\nwhere \\(A\\) is a measurable subset of \\(E\\). This notation specifies a transition law even when conditioning on \\(X_s=x\\) requires interpretation through a regular conditional distribution. The Chapman–Kolmogorov equations state that\n\n\\[P_{s,t}(x,A) = \\int_E P_{r,t}(z,A)\\,P_{s,r}(x,dz), \\qquad s\\leq r\\leq t.\\]\n\nThey express the fact that evolution over a time interval can be decomposed through an intermediate time.\n\nFor a suitable test function \\(f\\), define the transition operator\n\n\\[(U(s,t)f)(x) = \\int_E f(y)\\,P_{s,t}(x,dy).\\]\n\nThus \\(U(s,t)f\\) is a conditional expectation of the future observable \\(f(X_t)\\), viewed as a function of the present state. Composition satisfies\n\n\\[U(s,t)=U(s,r)U(r,t).\\]\n\nIn the time-homogeneous case, \\(U(s,t)=T_{t-s}\\), and the family \\(T_t\\) is a Markov semigroup.\n\nThe infinitesimal generator at time \\(t\\) is formally\n\n\\[L_t f = \\lim_{h\\downarrow0} \\frac{U(t,t+h)f-f}{h}.\\]\n\nThe limit is defined on an appropriate domain of functions, with a specified mode of convergence. In finite dimensions it is simply a matrix derivative; for diffusions it is usually an unbounded differential operator.\n\nDifferentiating the transition composition law gives the two operator equations\n\n\\[\\frac{\\partial}{\\partial t}U(s,t) = U(s,t)L_t\\]\n\nand\n\n\\[\\frac{\\partial}{\\partial s}U(s,t) = -L_sU(s,t),\\]\n\non suitable domains. The first varies the terminal time \\(t\\); the second varies the starting time \\(s\\). The placement of \\(L_t\\) matters because operators at different times need not commute.\n\nIf \\(\\mu_t\\) denotes the law of \\(X_t\\), its evolution can be written weakly as\n\n\\[\\frac{d}{dt}\\int_E f\\,d\\mu_t = \\int_E L_t f\\,d\\mu_t.\\]\n\nThis is the forward equation in a form that does not require a probability density. Writing \\(L_t^*\\) for the corresponding adjoint action on measures gives \\(\\partial_t\\mu_t=L_t^*\\mu_t\\). Conversely, for a fixed terminal time \\(T\\),\n\n\\[u(s,x)=\\mathbb E[f(X_T)\\mid X_s=x]\\]\n\nsatisfies \\(\\partial_su+L_su=0\\), with terminal condition \\(u(T,x)=f(x)\\). The two equations therefore describe a single transition mechanism acting on two different objects: distributions and observables."
    },
    {
      "heading": "Continuous-time chains and probability flow",
      "text": "For a finite-state chain, let \\(Q(t)=(q_{ij}(t))\\) be the rate matrix, using the convention that \\(q_{ij}\\) is the rate of transition from \\(i\\) to \\(j\\). Its off-diagonal entries are nonnegative, and\n\n\\[q_{ii}(t)=-\\sum_{j\\ne i}q_{ij}(t).\\]\n\nFor sufficiently regular rates,\n\n\\[\\Pr(X_{t+h}=j\\mid X_t=i) = \\delta_{ij}+q_{ij}(t)h+o(h).\\]\n\nThe diagonal term records loss from a state, while each off-diagonal term records transfer to another state.\n\nLet \\(P(s,t)\\) be the transition matrix, with entries \\(p_{ij}(s,t)\\). The forward and backward equations are\n\n\\[\\frac{\\partial}{\\partial t}P(s,t)=P(s,t)Q(t),\\]\n\n\\[\\frac{\\partial}{\\partial s}P(s,t)=-Q(s)P(s,t),\\]\n\nwith \\(P(s,s)=I\\). Entrywise, the forward equation is\n\n\\[\\frac{\\partial p_{ij}}{\\partial t}(s,t) = \\sum_k p_{ik}(s,t)q_{kj}(t).\\]\n\nIt balances all ways of entering or leaving the destination state \\(j\\). The backward equation instead resolves the first transition away from the initial state \\(i\\).\n\nFor time-homogeneous rates, put \\(P(t)=P(0,t)\\). Then\n\n\\[P'(t)=P(t)Q=QP(t), \\qquad P(t)=e^{tQ},\\]\n\nwhere \\(e^{tQ}\\) is the matrix exponential. The equality of the two products follows because \\(P(t)\\) is a function of the same fixed matrix \\(Q\\); it does not imply that arbitrary time-dependent rate matrices commute.\n\nFor example, a two-state chain with rates \\(\\alpha>0\\) from state \\(0\\) to state \\(1\\), and \\(\\beta>0\\) in the reverse direction, has\n\n\\[Q= \\begin{pmatrix} -\\alpha&\\alpha\\\\ \\beta&-\\beta \\end{pmatrix}.\\]\n\nStarting in state \\(0\\),\n\n\\[p_{01}(t) = \\frac{\\alpha}{\\alpha+\\beta} \\left(1-e^{-(\\alpha+\\beta)t}\\right).\\]\n\nThe approach to equilibrium occurs on the characteristic time scale \\(1/(\\alpha+\\beta)\\).\n\nA distribution written as a row vector satisfies \\(\\pi'(t)=\\pi(t)Q(t)\\). With column-vector distributions, the same equation becomes \\(\\pi'(t)=Q(t)^{\\mathsf T}\\pi(t)\\). Many apparent discrepancies in formulas result only from this convention.\n\nFor countably many states, the equations involve infinite sums and require additional convergence and nonexplosion conditions. A birth–death chain with birth rates \\(\\lambda_n\\) and death rates \\(\\mu_n\\) satisfies\n\n\\[p_n' = \\lambda_{n-1}p_{n-1} + \\mu_{n+1}p_{n+1} - (\\lambda_n+\\mu_n)p_n,\\]\n\nwith nonexistent terms omitted at the boundary. This gain–loss form is often called a master equation. Unlike the finite-state case, its formal algebra alone does not establish conservation of total probability."
    },
    {
      "heading": "Diffusions and the Fokker–Planck equation",
      "text": "Consider an Itô stochastic differential equation in \\(\\mathbb R^d\\),\n\n\\[dX_t=b(t,X_t)\\,dt+\\sigma(t,X_t)\\,dW_t,\\]\n\nwhere \\(W_t\\) is an \\(m\\)-dimensional Wiener process. Define the local covariance matrix\n\n\\[a(t,x)=\\sigma(t,x)\\sigma(t,x)^{\\mathsf T}.\\]\n\nFor a twice differentiable test function, the generator is\n\n\\[L_tf(x) = \\sum_{i=1}^d b_i(t,x)\\frac{\\partial f}{\\partial x_i} + \\frac12\\sum_{i,j=1}^d a_{ij}(t,x) \\frac{\\partial^2f}{\\partial x_i\\partial x_j}.\\]\n\nThe first-order term describes systematic drift; the second-order term describes the accumulation of random fluctuations. The factor \\(1/2\\) comes from the second-order term in Itô’s formula.\n\nIf the law has a sufficiently regular probability density \\(\\rho(t,x)\\) relative to Lebesgue measure, its forward equation is\n\n\\[\\frac{\\partial\\rho}{\\partial t} = -\\sum_i\\frac{\\partial}{\\partial x_i}(b_i\\rho) + \\frac12\\sum_{i,j} \\frac{\\partial^2}{\\partial x_i\\partial x_j}(a_{ij}\\rho).\\]\n\nThis is the Fokker–Planck equation. For a transition density \\(p(s,x;t,y)\\), the derivatives in the forward equation act on the terminal variables \\(t,y\\):\n\n\\[\\partial_t p=L_{t,y}^*p.\\]\n\nThe backward equation acts instead on the starting variables:\n\n\\[\\partial_s p+L_{s,x}p=0.\\]\n\nWith spatially varying coefficients, the forward operator is generally not obtained merely by replacing \\(f\\) by \\(\\rho\\) in the generator: derivatives also act on the coefficients.\n\nFor standard Brownian motion, \\(b=0\\) and \\(a=I\\). The forward equation becomes the heat equation,\n\n\\[\\partial_t\\rho=\\frac12\\Delta\\rho,\\]\n\nwhere \\(\\Delta\\) is the Laplacian. Its transition density is\n\n\\[p(s,x;t,y) = \\frac{1}{[2\\pi(t-s)]^{d/2}} \\exp\\left(-\\frac{|y-x|^2}{2(t-s)}\\right).\\]\n\nAt \\(t=s\\), the initial condition is a point mass at \\(x\\), not an ordinary density function. The density approaches that point mass weakly as \\(t\\downarrow s\\).\n\nA useful nontrivial example is the Ornstein–Uhlenbeck process,\n\n\\[dX_t=-\\theta X_t\\,dt+\\sigma\\,dW_t, \\qquad \\theta>0.\\]\n\nGiven \\(X_0=x\\), its mean is \\(xe^{-\\theta t}\\) and its variance is\n\n\\[\\frac{\\sigma^2}{2\\theta}(1-e^{-2\\theta t}).\\]\n\nIts forward equation,\n\n\\[\\partial_t\\rho = \\theta\\,\\partial_x(x\\rho) + \\frac{\\sigma^2}{2}\\partial_{xx}\\rho,\\]\n\ncombines concentration toward the origin with diffusion away from concentrated states. The limiting density is normal with variance \\(\\sigma^2/(2\\theta)\\)."
    },
    {
      "heading": "Adjointness, probability current, and boundaries",
      "text": "The relation between the equations is a form of duality. For a distribution \\(\\mu\\) and an observable \\(f\\),\n\n\\[\\int_E f\\,d(U(s,t)^*\\mu) = \\int_E U(s,t)f\\,d\\mu.\\]\n\nThe left-hand side evolves the distribution and then evaluates the observable; the right-hand side first computes the observable’s expected future value and then averages over the initial distribution. Equality is a consequence of conditional expectation.\n\nFor differential operators, the adjoint formula follows from integration by parts. On a domain \\(D\\),\n\n\\[\\int_D \\rho Lf\\,dx = \\int_D fL^*\\rho\\,dx + \\text{boundary terms}.\\]\n\nThe differential expression for \\(L^*\\) does not by itself determine the evolution. The domain of the operator, including its boundary conditions, is equally important. Two processes may have the same interior drift and diffusion coefficients but behave differently when they reach the boundary.\n\nThe forward diffusion equation can be written as a conservation law,\n\n\\[\\partial_t\\rho+\\nabla\\cdot J=0,\\]\n\nwhere the probability current is\n\n\\[J_i = b_i\\rho - \\frac12\\sum_j\\partial_{x_j}(a_{ij}\\rho).\\]\n\nThe divergence theorem then gives\n\n\\[\\frac{d}{dt}\\int_D\\rho\\,dx = -\\int_{\\partial D}J\\cdot n\\,dS,\\]\n\nwhere \\(n\\) is the outward normal. Changes in total probability inside \\(D\\) are therefore governed by flux across its boundary.\n\nFor reflected Brownian motion in a smooth bounded domain, the backward operator has the Neumann condition \\(\\partial_n f=0\\); the corresponding forward density has zero normal current. More general reflected diffusions require boundary conditions adapted to the specified reflection direction. An arbitrary anisotropic diffusion cannot automatically be assigned the same pair of boundary conditions as isotropic Brownian motion.\n\nFor a process killed when it exits \\(D\\), the killed backward semigroup has homogeneous Dirichlet boundary conditions. Under standard regularity and nondegeneracy assumptions, its interior forward density also vanishes at the absorbing boundary, and its integral decreases as probability is lost through absorption. If the process is instead retained at absorbing boundary states, the full law remains a probability measure but may acquire singular mass on the boundary. An interior density alone then represents only part of the law.\n\nBoundary conditions are unnecessary only when the state space genuinely has no boundary or when appropriate behavior at infinity is built into the function spaces. Decay, integrability, and nonexplosion assumptions can play the role of boundary information on unbounded domains. The pairing between forward and backward solutions remains valid only when the relevant boundary contributions vanish or are explicitly included."
    },
    {
      "heading": "Existence, uniqueness, and explosion",
      "text": "Writing down a generator does not automatically produce a unique Markov process or a unique solution of its Kolmogorov equations. Three questions must be distinguished: whether a stochastic process exists, whether its law is unique, and whether the differential equation has a unique solution in a specified class. Connections among these questions are powerful, but they depend on coefficient assumptions, operator domains, and the meaning of “solution.”\n\nFor finite-state chains with bounded, sufficiently regular time-dependent rates, standard linear ordinary differential equation theory provides a unique transition matrix. Positivity and preservation of row sums follow from the structure of the rate matrix. With infinitely many states, rates can grow so rapidly that infinitely many jumps occur in finite time.\n\nA standard example is a pure-birth chain with positive rates \\(\\lambda_n\\). Its successive holding times are independent exponential random variables with means \\(1/\\lambda_n\\). If\n\n\\[\\sum_{n=0}^{\\infty}\\frac{1}{\\lambda_n}<\\infty,\\]\n\nthe chain makes infinitely many jumps in finite time almost surely. For instance, \\(\\lambda_n=(n+1)^2\\) produces explosion. The minimal transition law loses mass from the original state space after explosion. Formal row-sum identities for the rate matrix do not prevent this loss, because infinite summation and differentiation cannot be interchanged without justification.\n\nFor stochastic differential equations on \\(\\mathbb R^d\\), global Lipschitz continuity in the state variable together with suitable linear-growth bounds gives a standard sufficient framework for a unique nonexplosive strong solution. These conditions are not necessary. Local conditions combined with a suitable Lyapunov function often suffice, while superlinear outward drift can lead to explosion.\n\nClassical solutions of the diffusion equations require stronger regularity than existence of the underlying process. Smooth coefficients and uniform ellipticity provide a well-developed theory of smooth transition densities. Degenerate diffusions may nevertheless have smooth densities when the drift and diffusion directions satisfy appropriate bracket conditions; other degenerate processes have singular laws and no full-dimensional density.\n\nA weak solution of the forward equation is formulated through test functions:\n\n\\[\\int f\\,d\\mu_t-\\int f\\,d\\mu_0 = \\int_0^t\\int L_sf\\,d\\mu_s\\,ds.\\]\n\nThis formulation accommodates point-mass initial data and distributions that lack smooth densities.\n\nThe martingale problem provides a further bridge between operators and processes. It requires that, for suitable \\(f\\),\n\n\\[f(X_t)-f(X_0)-\\int_0^tL_sf(X_s)\\,ds\\]\n\nbe a martingale. Well-posedness of this problem identifies the stochastic law associated with the operator. It is especially useful when coefficients are too irregular for a direct classical PDE treatment."
    },
    {
      "heading": "Backward equations for observables and stopping problems",
      "text": "The backward equation is particularly effective when the desired quantity is a function of the starting state rather than a complete distribution. For a terminal payoff \\(g\\),\n\n\\[u(t,x)=\\mathbb E_{t,x}[g(X_T)]\\]\n\nsatisfies\n\n\\[\\partial_tu+L_tu=0, \\qquad u(T,x)=g(x).\\]\n\nAlthough called “backward,” this is not an unstable reversal of a realized random trajectory. It computes how the expected terminal payoff depends on the current state and remaining time. In a time-homogeneous problem, changing variables to the time remaining, \\(r=T-t\\), converts it to a forward initial-value equation \\(\\partial_rv=Lv\\).\n\nA central tool is Dynkin’s formula, which, under suitable integrability conditions, gives\n\n\\[\\mathbb E_x[f(X_\\tau)]-f(x) = \\mathbb E_x\\left[\\int_0^\\tau Lf(X_s)\\,ds\\right]\\]\n\nfor an appropriate stopping time \\(\\tau\\). This identity converts questions about random paths into boundary-value problems for the generator.\n\nFor example, let \\(\\tau_D\\) be the first exit time from a domain \\(D\\), and let \\(h\\) prescribe a payoff on the boundary. Then\n\n\\[u(x)=\\mathbb E_x[h(X_{\\tau_D})]\\]\n\ntypically solves\n\n\\[Lu=0\\quad\\text{in }D, \\qquad u=h\\quad\\text{on }\\partial D.\\]\n\nFor Brownian motion, this becomes Laplace’s equation. Harmonic functions consequently describe expected boundary values sampled by Brownian exit locations.\n\nThe expected exit time,\n\n\\[m(x)=\\mathbb E_x[\\tau_D],\\]\n\nsatisfies\n\n\\[Lm=-1\\quad\\text{in }D, \\qquad m=0\\quad\\text{on }\\partial D,\\]\n\nprovided the expectation is finite and the representation is justified. For standard one-dimensional Brownian motion in \\((0,\\ell)\\),\n\n\\[\\frac12m''=-1, \\qquad m(0)=m(\\ell)=0,\\]\n\nso \\(m(x)=x(\\ell-x)\\). In the same interval, the probability of exiting at \\(\\ell\\) rather than at \\(0\\) is \\(x/\\ell\\).\n\nThe Feynman–Kac formula extends the backward equation to running costs and killing or discounting. If\n\n\\[u(t,x)= \\mathbb E_{t,x}\\left[ e^{-\\int_t^T c(r,X_r)\\,dr}g(X_T) + \\int_t^T e^{-\\int_t^s c(r,X_r)\\,dr} r_0(s,X_s)\\,ds \\right],\\]\n\nthen, under suitable hypotheses,\n\n\\[\\partial_tu+L_tu-cu+r_0=0, \\qquad u(T,\\cdot)=g.\\]\n\nThe coefficient \\(c\\) weights survival or discounting, while \\(r_0\\) is an accumulated reward rate. Optimizing over a controlled family of generators leads to nonlinear Hamilton–Jacobi–Bellman equations rather than the linear Kolmogorov equation."
    },
    {
      "heading": "Stationarity, detailed balance, and time reversal",
      "text": "A stationary distribution is unchanged by the transition dynamics. For a time-homogeneous chain, a stationary row vector \\(\\pi\\) satisfies\n\n\\[\\pi Q=0, \\qquad \\sum_i\\pi_i=1.\\]\n\nFor a diffusion with stationary density \\(\\rho_\\infty\\), the corresponding condition is\n\n\\[L^*\\rho_\\infty=0, \\qquad \\int\\rho_\\infty\\,dx=1,\\]\n\ntogether with the appropriate boundary behavior. These equations identify candidate equilibrium laws, but do not by themselves prove convergence to equilibrium. Irreducibility, recurrence, and mixing properties determine whether a stationary law exists uniquely and attracts other initial distributions.\n\nA stronger condition is detailed balance. For a chain, it requires\n\n\\[\\pi_iq_{ij}=\\pi_jq_{ji}\\]\n\nfor every pair of states. Each stationary probability flux is balanced by its reverse flux. Summing this relation over one index gives stationarity, but stationarity alone does not imply detailed balance. A directed cycle can carry a nonzero stationary circulating current while preserving its distribution.\n\nFor a diffusion, vanishing stationary probability current is the familiar counterpart of detailed balance under standard assumptions. Consider\n\n\\[dX_t=-\\nabla V(X_t)\\,dt+\\sqrt{2D}\\,dW_t, \\qquad D>0.\\]\n\nIf\n\n\\[Z=\\int_{\\mathbb R^d}e^{-V(x)/D}\\,dx<\\infty,\\]\n\nthen\n\n\\[\\rho_\\infty(x)=Z^{-1}e^{-V(x)/D}\\]\n\nhas zero current. This connects the forward equation to equilibrium statistical mechanics. Nonzero divergence-free stationary currents instead describe stationary systems without detailed balance.\n\nThe backward Kolmogorov equation must be distinguished from the dynamics of a process observed in reverse time. For a stationary chain with positive stationary probabilities, the reversed generator has off-diagonal rates\n\n\\[q^{\\mathrm{rev}}_{ij} = \\frac{\\pi_jq_{ji}}{\\pi_i}.\\]\n\nThe reversed chain has the same generator as the original precisely when detailed balance holds.\n\nFor a diffusion with smooth positive density \\(\\rho(t,x)\\), define \\(Y_s=X_{T-s}\\). Under suitable regularity assumptions, its reversed drift is\n\n\\[b_i^{\\mathrm{rev}}(s,x) = -b_i(T-s,x) + \\frac{1}{\\rho(T-s,x)} \\sum_j \\partial_{x_j} \\left(a_{ij}(T-s,x)\\rho(T-s,x)\\right),\\]\n\nwith reversed local covariance \\(a(T-s,x)\\). When \\(a\\) is spatially constant, the correction involves \\(a\\nabla\\log\\rho\\).\n\nThis density-dependent correction shows why reversing a stochastic process is not achieved by merely negating its drift. By contrast, the backward Kolmogorov equation continues to use the original generator: “backward” refers to the starting-time variable, not to a different physical process."
    },
    {
      "heading": "Numerical solution and structural preservation",
      "text": "For finite chains, direct evaluation of \\(e^{tQ}\\) is effective when the state space is modest. Large models usually exploit a sparse matrix representation and compute the action of the exponential on a vector rather than forming the full transition matrix. This distinction can reduce both storage and computational work substantially.\n\nA probability-preserving method for a homogeneous finite-state chain is uniformization. Choose\n\n\\[\\lambda\\geq\\max_i(-q_{ii}), \\qquad R=I+\\frac{Q}{\\lambda}.\\]\n\nThen \\(R\\) is a discrete-time transition matrix and\n\n\\[e^{tQ} = e^{-\\lambda t} \\sum_{n=0}^{\\infty} \\frac{(\\lambda t)^n}{n!}R^n.\\]\n\nThe number of potential updates has a Poisson distribution with mean \\(\\lambda t\\), and some updates leave the state unchanged. Truncating the series yields an explicit error bound from the omitted Poisson tail. For unbounded exit rates on an infinite state space, a single finite uniformization rate may not exist.\n\nDiffusion equations are treated with finite differences, finite-volume methods, finite elements, and spectral approximations. Forward solvers should preserve nonnegativity and total probability whenever the model conserves mass. A scheme may be accurate in a conventional norm yet produce negative densities if its discrete operator does not respect the generator structure.\n\nFinite-volume methods approximate probability fluxes across cell faces, making conservation explicit. Upwind treatment of drift can help retain positivity but introduces numerical diffusion. Centered approximations can be more accurate for smooth solutions while generating oscillations in drift-dominated regimes. Explicit time-stepping typically requires restrictive step sizes; implicit methods handle stiffness more effectively but require linear solves.\n\nThe forward–backward duality also provides a design principle for discretization. If a discrete backward generator acts on observables, its transpose, interpreted with the relevant grid or quadrature weights, should propagate discrete probability masses. This makes the discrete identity between expected payoffs and evolved distributions hold exactly. The distinction between a vector of cell masses and a vector of density values is important on nonuniform meshes.\n\nIn high dimensions, deterministic grids encounter the curse of dimensionality. Monte Carlo methods instead estimate backward expectations from simulated paths, with the usual sampling error decreasing as the inverse square root of the number of independent samples when variance is finite. Simulation introduces time-discretization error as well as sampling error. Estimating a rare transition probability or a full high-dimensional density can remain difficult even when individual paths are inexpensive to generate."
    },
    {
      "heading": "Historical development and scientific applications",
      "text": "The modern general theory is associated with Andrey Kolmogorov, whose 1931 paper Über die analytischen Methoden in der Wahrscheinlichkeitsrechnung developed analytical methods for Markov transition functions. Its treatment brought forward and backward differential equations into a common probabilistic framework. The work preceded his 1933 axiomatic formulation of probability and addressed the relation between local transition behavior and global evolution.\n\nImportant diffusion equations had already appeared in physics. Albert Einstein’s 1905 analysis of Brownian motion related random molecular motion to diffusion, while Marian Smoluchowski developed related kinetic descriptions. Adriaan Fokker in 1914 and Max Planck in 1917 obtained evolution equations in particular physical settings. These strands explain the overlapping terminology: “Fokker–Planck equation” emphasizes the forward evolution of a diffusion density, whereas “Kolmogorov equations” places forward and backward equations within general Markov-process theory.\n\nLater work on stochastic integration, semigroups, boundary classification, and martingale problems clarified when formal differential operators determine actual stochastic dynamics. Itô’s stochastic calculus supplied a systematic route from stochastic differential equations to generators. Operator methods connected these probabilistic constructions with functional analysis, while weak formulations extended the equations beyond smooth densities and classical solutions.\n\nApplications often divide naturally according to which direction is computationally useful. In chemical reaction networks, the state records molecule counts, and the forward chemical master equation tracks their distribution. A reaction with a given propensity transfers probability from one count vector to another. Diffusion approximations replace the jump generator by a second-order differential operator when an appropriate scaling and truncation are justified.\n\nIn population models, forward equations describe distributions of population sizes, allele frequencies, or extinction states. Backward equations are often more economical for fixation probabilities and expected extinction times because they solve directly for a quantity indexed by the initial population. In queueing models, the same distinction separates transient queue-length distributions from waiting-time and hitting-time calculations.\n\nIn mathematical finance, pricing expectations lead to backward equations with discounting. For a geometric Brownian price process under a risk-neutral model, the generator has drift term \\(rx\\partial_x\\) and diffusion term \\(\\tfrac12\\sigma^2x^2\\partial_{xx}\\); adding the discount term produces the Black–Scholes pricing equation. The forward equation instead propagates the model’s price distribution.\n\nThe distinction also appears in contemporary diffusion-based generative modeling. A prescribed noising process evolves a data distribution through a forward Fokker–Planck equation. Generating samples by a reversed diffusion requires the reversed drift and a density-score term, rather than simply solving the backward Kolmogorov equation. Despite related terminology, distribution propagation, observable propagation, and stochastic time reversal remain mathematically distinct operations."
    }
  ],
  "links": [
    "stochastic-process",
    "markov-property",
    "continuous-time-markov-chain",
    "chapman-kolmogorov-equations",
    "conditional-expectation",
    "matrix",
    "transition-matrix",
    "matrix-exponential",
    "stochastic-differential-equation",
    "wiener-process",
    "probability-density-function",
    "heat-equation",
    "laplace-operator",
    "integration-by-parts",
    "divergence-theorem",
    "ordinary-differential-equation",
    "lipschitz-continuity",
    "weak-solution",
    "martingale",
    "laplace-equation",
    "feynman-kac-formula",
    "stationary-distribution",
    "detailed-balance",
    "sparse-matrix",
    "poisson-distribution",
    "finite-difference-method",
    "finite-element-method",
    "curse-of-dimensionality",
    "monte-carlo-method",
    "andrey-kolmogorov",
    "functional-analysis"
  ],
  "sources": [],
  "license": "https://creativecommons.org/licenses/by/4.0/"
}