Genealogies and increasing propagation of chaos for Feynman-Kac and genetic models. (Q1872412)

This quite technical paper contains proofs of statements about convergence and asymptotic behaviour of some specific empirical measures as they typically arise in some forms of theoretical filtering problems where Markovian ideas predominate. Generalizations are at the forefront and estimates are in terms of entropy and \(L_\alpha\) norms, \(\alpha\geq 1\). It turns out that such problems can be expressed and tackled with ideas from genetic algorithms and it is shown that previous methods can be adapted so that one can trace at each time the complete genealogy of each individual. The editing has been unusually poor. The worst offence occurs at the end of Theorem 3.1 which incidentally is given, at the beginning of the section, as the proof of Theorem 1.1 (though there are two different statements). It says: ``This ends the proofs of the theorem. The proof of Lemma 3.4 is now complete''. The proof of that lemma (3.4) follows the latter statement! Also it is not always reasonable to assume that sonar statistics are known as underwater signals can be highly transient and ``noiselike''.