Special ACM Seminar
Annenberg 213
Neural Signature Kernels as Infinite-Width Limits of Neural Controlled Differential Equations
Cristopher Salvi,
Faculty of Natural Sciences,
Department of Mathematics,
Imperial College - London,
Motivated by the paradigm of reservoir computing, I will consider randomly initialized neural controlled differential equations and show that in the infinite-width limit and under proper rescaling of the vector fields, these neural architectures converge weakly to Gaussian processes indexed on path-space and with covariances satisfying certain PDEs varying according to the choice of activation function. In the special case where the activation function is the identity, the equation reduces to a linear PDE and the limiting kernel agrees with the original signature kernel.
This is based on joint work with Nicola M. Cirone and Maud Lemercier.
For more information, please contact Diana Bohler by phone at 6263951768 or by email at [email protected].
Event Series
Special Seminars in Applied Mathematics