SVI tends to break, when the (local) optimum is reached. Why this happens, I don't know. It would be good to stop the loop, when the change in loss value becomes very small. This would a) speed up SVI and b) reduce non-converging chains. Also it allows greater control over SVI, for instance kicking out infinity simulations.
SVI tends to break, when the (local) optimum is reached. Why this happens, I don't know. It would be good to stop the loop, when the change in loss value becomes very small. This would a) speed up SVI and b) reduce non-converging chains. Also it allows greater control over SVI, for instance kicking out infinity simulations.