Recent study approach with Ambient-Latent Space resembles a new possibility of simulating dynamics with more accuracy. This paper aims to suggest idea also by adding temporary dimension(like in case of SVM) in learning given dynamics by optimization in additional to the recent study.