Complexity and irreversibility in stochastic analysis (Q1600491)

The author discusses complexity of random phenomena which can be expressed as functionals of white noise \(\dot B(t)=\frac{dB(t)}{dt}\). This white noise is realized as the time derivative of a Brownian motion \( B(t)\), and the collection \(\{ \dot B(t)\} \) is a system of idealized elemental variables. The author introduces first the notion of multiplicity in time evolution, which describes how much the phenomena are complex. In fact many significant random complex systems are represented in a causal manner by linear or nonlinear functions of \(B(t)\), with multiplicity, where the passage of time is always taken into account. For example let \( X(t)\) be a Gaussian process with mean \(E(X(t))=0.\) Assume that \(X(t)\) is separable and has unit multiplicity in the time domain. Then there exists a white noise \(\dot B(t)\) such that \( X(t)=\int^{t}F(t,u)\dot B(t)du\) where \(F(t,u)\) is a nonrandom kernel function. In addition \(\{ X(u),u\leq t\} \) has the same information as \(\{ \dot B(t) ,u\leq t\} \) for every \(t.\) In order to establish the causal calculus of complex systems of the above form of a stochastic process, the author generalizes the notion of the multiplicity, namely, a one parameter unitary group \(\{ U(t),t\in R\} \), acting on the space of white noise functionals and representing the time propagation and verifying the relation \(U(t)\dot B(t)=\dot B(t+s)\). The system \(\{ H(f_{n});n=1,2,\dots \} \), where \(H(f)=\) span\(\{ U(t)f,t\in R\} \), is an orthogonal system and the complex system in question is the direct sum of those cyclic subspaces. The number of these cyclic subspaces is the multiplicity in the general sense, and the multiplicity is one of the measurements that express the degree of complexity. Finally the author introduces the notion of reversibility and irreversibility of random evolutional phenomena and gives technical conditions that a bridged Gaussian process of the form \(X(t)=\int_{0}^{t}R(t,u)\dot B(t) du \), where \(R(t,u)=\sum_{k=1}^{k=N}a_{k}\frac{(1-t)^{k}}{(1-u)^{k}}\) is a Riemannian function, is a reversible process.