Keynotes | K&Ts | GACs | Talks | Posters | Search

Poster A in Poster Session A: Tuesday, August 4, 9:30 – 11:15 am, Kimmel Center, Shorin & Rosenthal Rooms

Unifying discrimination and generation in one architecture

Imran Thobani¹, Klemen Kotar¹, Jared Watrous¹, Andreas S. Tolias², Daniel LK Yamins¹; ¹Stanford University, ²Baylor College of Medicine

Presenter: Imran Thobani

The primate visual system exhibits both discriminative and generative abilities. For example, we can not only recognize a dog when we see it (a discriminative ability), but we can also imagine a dog with our eyes closed (a generative ability). However, the current best models of the visual system explain only discriminative abilities, such as object recognition. Here, we present a model that can flexibly perform discrimination and generation tasks, including hybrid tasks that combine aspects of both, providing a concrete model for investigating how these seemingly different computations could be unified in a single neural architecture.

Topic Area: Computational Models of Vision & Visual Cortex