Functional Generative Encoders

Functions and Differentials

25/02/2022

Proposal: Finite Element Methods

Observed data can be represented by points, a measurable change between points is a difference. The path that routes different points together can be considered a function. Differentials and derivatives are used to quantify the change in step sized changes using finite differences. An approach to extend Generative Encoders to be more functional could be treating the observed data as points, the learned transform as a function, and treating perturbations between observed points as elements with a finite difference.