@reidatcheson In terms of cutting a graph, I would use heuristic like "filling" the edges with some filler, then would apply temperature and shortly after that pressure. Iterative solver would eventually frac a stracture along the weak edges, but temperature would glue the strong ones together
@reidatcheson It's vaguely related, stumbled upon this while working on symbolically constructing matrices... "Twelve Proofs of the Markov Inequality
Aleksei Shadrin".
@danbgoldman@docmilanfar I love this version, the most correct representation I've read. Reminds me on one of the articles about style transfer (can't find now ๐). However, it's just one interpretation/implementation of diffusion model. @docmilanfar has similar, more probabilistic interpretation.
@kolibril13 Just use @Panel_org - well maintained, provides a rich ecosystem and more deployment options, and integrates with third-party than anything else.
Recently read (audio book) The Ants by Boris Vian... About war, about its casual ugly side... Impressed is not the right word... appalled, shocked, stun, shattered by truth... #powerofwords#Literature#StopTheWar
@science_dot MPI is good for speed (productivity and performance) in HPC. However, I'm waiting for the moment when I can add and remove ranks (by allocating and deallocating machines) dynamically without restarting my application from the checkpoint. Dynamically scaling in and out for short.