Nonlinear transformations on the Wiener space (Q1317238)

The author studies transformations on an abstract Wiener space \((E, H, \mu)\) which have the form \[ E\ni w\to S(w)= w- \int^ T_ 0 \varphi_ t (w) Z(dt)\in E, \] where \(\varphi_ t(w)\) is a scalar function on \([0,T]\times E\) and \(Z\) is an orthogonal \(H\)-valued measure. Assuming that \(\varphi\in \mathbb{L}^{1,2}\) and that the Malliavin derivative \(D\varphi\) of \(\varphi\) is such that \(| D\varphi (w)|_{H,S} <1\), \(\mu(dw)\)-a.e., the author applies a special factorization technique developed by Gokhberg and Krejn (1970) to express explicitly the Carleman-Fredholm determinant which occurs in Kusuoka's formula (1982) of the density \(R_ \varphi\) of the image of \(\mu\) under \(S\) w.r.t. \(\mu\) as exponential of the product of two integral kernels. This approach works also in cases where \(S\) is not invertible. The main result, stated under the assumptions presented above, constitutes an anticipative analogon of the Cameron-Martin-Girsanov formula: if \(R_ \varphi\) has the expectation 1 under \(\mu\), then \((R_ \varphi\mu)\cdot S^{-1} =\mu\). This generalizes also a paper of the reviewer (1991) who has determined explicitly the integral kernels from the factorization of the Carleman-Fredholm determinant under the assumption \(\text{ess sup}_{w\in E} | D_ \varphi (w)|_{H\cdot S} <1\).