Research Update

I just posted an article that goes over the context of my research and how 3-way autoencoders fit in.  I worked on pulling out some of the information from the papers indicated in that article into my research bible. In this article I also drafted an architecture for a cross-modal/cross-domain transfer learning framework. I need to define cross-modality.  In doing so, I need a good definition of task and task domain.  I'll probably be using the definition from: [1] K. D. Feuz, “Transfer Learning across Feature-Rich Heterogeneous Feature Spaces via Feature-Space Remapping ( FSR ),” ACM Trans. Intell. Syst. Technol., vol. V, no. 212, 2014. This paper defines domain $latex D$ as a two-tuple $(\chi, P(X))$ where $latex \chi$ is the feature space of $latex D$, and $latex P(X)$ is a marginal distribution…