A certain class of infinite dimensional diffusion processes arising in population genetics (Q579771): Difference between revisions

Let X denote the set of all \(x=(x_ i)\in \ell^ 1\) with \(\| x\| =1\) and \(x\geq 0\), equipped with the weak topology. Consider the differential operator \[ L=2^{-1}\sum_{i,j}a_{ij}(x)\partial^ 2/\partial x_ i\partial x_ j+\sum_{i}b_ i(x)\partial /\partial x_ i \] on X, where \(a_{ij}(x)=(x_ i\beta_ i\delta_{ij}+x_ ix_ j)(\sum_{k}x_ k\beta_ k-\beta_ i-\beta_ j)\) where \(\beta =(\beta_ i)\in \ell^{\infty}\), \(\beta\geq 0\) and \(\delta_{ij}= Kronec\ker\) symbol. Let \(W=C({\mathbb{R}}_+,X)\) denote the space of continuous functions on \({\mathbb{R}}_+\) with values in X with the usual filtration (\({\mathcal F}_ t)\) given by the cylinder sets (up to time t). A diffusion on X for the operator L is defined as a strongly Markovian family \((P_ x)_{x\in X}\) of probability measures on W with the properties (a) \(P_ x\{w:\) \(w(0)=x\}=1\) and \[ (b)\quad (f(w(t))- f(w(0))-\int^{t}_{0}Lf(w(s))ds)_{t\geq 0} \] is a \((P_ x,{\mathcal F}_ t)\)-martingale for every \(C^ 2\)-function on X depending only on finite many coordinates. The main result is an existence and uniqueness theorem for these diffusions on X under suitable assumptions on the drift b. The above martingale problem is solved by the method of stochastic differential equations.