Efficient inverse graphics in biological face processing.

Vision not only detects and recognizes objects, but performs rich inferences about the underlying scene structure that causes the patterns of light we see. Inverting generative models, or "analysis-by-synthesis", presents a possible solution, but its mechanistic implementations have typically been too slow for online perception, and their mapping to neural circuits remains unclear. Here we present a neurally plausible efficient inverse graphics model and test it in the domain of face recognition. The model is based on a deep neural network that learns to invert a three-dimensional face graphics program in a single fast feedforward pass. It explains human behavior qualitatively and quantitatively, including the classic "hollow face" illusion, and it maps directly onto a specialized face-processing circuit in the primate brain. The model fits both behavioral and neural data better than state-of-the-art computer vision models, and suggests an interpretable reverse-engineering account of how the brain transforms images into percepts.

Topographical functional connectivity patterns exist in the congenitally, prelingually deaf.

Congenital deafness causes large changes in the auditory cortex structure and function, such that without early childhood cochlear-implant, profoundly deaf children do not develop intact, high-level, auditory functions. But how is auditory cortex organization affected by congenital, prelingual, and long standing deafness? Does the large-scale topographical organization of the auditory cortex develop in people deaf from birth? And is it retained despite cross-modal plasticity? We identified, using fMRI, topographic tonotopy-based functional connectivity (FC) structure in humans in the core auditory cortex, its extending tonotopic gradients in the belt and even beyond that. These regions show similar FC structure in the congenitally deaf throughout the auditory cortex, including in the language areas. The topographic FC pattern can be identified reliably in the vast majority of the deaf, at the single subject level, despite the absence of hearing-aid use and poor oral language skills. These findings suggest that large-scale tonotopic-based FC does not require sensory experience to develop, and is retained despite life-long auditory deprivation and cross-modal plasticity. Furthermore, as the topographic FC is retained to varying degrees among the deaf subjects, it may serve to predict the potential for auditory rehabilitation using cochlear implants in individual subjects.