Poster
in
Workshop: Physics for Machine Learning
Learning Deformation Trajectories of Boltzmann Densities
Bálint Máté · François Fleuret
Abstract:
We introduce a training objective for continuous normalizing flows that can be used in the absence of samples but in the presence of an energy function. Our method relies on either a prescribed or a learnt interpolation of energy functions between the target energy and the energy function of a generalized Gaussian . The interpolation of energy functions induces an interpolation of Boltzmann densities and we aim to find a time-dependent vector field that transports samples along the family of densities. The condition of transporting samples along the family can be translated to a PDE between and and we optimize and to satisfy this PDE.
Chat is not available.