An overview of current techniques for self-supervised learning of spatio-temporal trajectories, a discussion on the constraints of these methods in terms of transferability, and a look at potential future research paths.
An overview of popular solutions for effective encoding of continuous features, classified into bucket encoding and fourier features.
An overview of the two technical routes toward adapting Diffusion Models in discrete feature space.
A brief overview of the Diffusion Models' framework, its implementation and applications.
A walk through of the AAAI accepted paper on pre-training contextual location embeddings from spatial-temporal trajectories using the proposed Context and Time aware Location Embedding (CTLE) model.