Infinitesimal generator (stochastic processes)
ID: infinitesimal-generator-stochastic-processes
The Infinitesimal generator describes the instantaneous action of a Markov transition semigroup on functions of the state. For a strongly continuous semigroup on a specified Banach space of functions, its generator isThe domain and the limit topology are part of the definition; different function spaces can give different operator domains. Pointwise generator formulas are often computed first on smooth test functions.
For a continuous-time Markov chain with rates ,For Brownian diffusion with variance , on smooth compactly supported functions. For a deterministic flow , the chain rule instead gives . These examples turn transition rules or local motion into an operator.
For suitable and integrability conditions, Dynkin's formula makes a martingale. Taking expectations yields evolution equations; stopping at a boundary gives equations such as for hitting probabilities and for mean first-passage times. The generator is useful because local drift, diffusion or jump rates encode global transition and boundary questions.
New to topics? Read the docs here!