I have collected some of my selected works for display. I will update the list from time to time.
We attempt to focus on sparse data using the modified radial basis function neural network in conjunction with density based sample generation algorithm.
We attempt to apply an incremental density clustering algorithm on a modified radial basis function neural network on classification tasks. By introducing the RBF parameters in the training process, the convergence rate of the training process is improved while sustaining a high level of accuracy in most of our experiments.
This is my dissertation for my undergraduate at the Chinese University of Hong Kong. In this project, I go through the process of using branching diffusion processes to solve Feynman-Kac formula (or parabolic differential equations in general). My contribution was to extend to the case where the nonlinearity term is an integro-differential operator and identify the correponding marked branching diffusion processes with jumps for a given PDE. I have also explored and verified sufficient conditions for such a correspondence in the smooth density case.