Abstract: We propose applying the reaction-diffusion equation on a graph to the function approximation for reinforcement learning. which realizes adaptive resolution according to the complexity of the ...

Abstract: Graph-based image analysis methods are increasingly being applied to diffusion ... function (ODF). To date, these integrals have been approximated numerically at high computational cost and ...

In this paper, we present a novel method based on time derivative graph diffusion (TIDE), with a learnable time parameter. Our approach allows us to adapt the spatial extent of diffusion across ...