Monday, January 30, 2017, 12:00 – 1:30 pm, in 489 Minor Hall
Graduate Student Seminar
Dylan Paiton, PhD Candidate (Olshausen Lab)
Emergence of foveal image sampling from learning to attend in visual scenes
We describe a neural attention model with a learnable retinal sampling lattice. The model is trained on a visual search task requiring the classification of an object embedded in a visual scene amidst background distractors using the smallest number of fixations. We explore the tiling properties that emerge in the model’s retinal sampling lattice after training. Specifically, we show that this lattice resembles the eccentricity dependent sampling lattice of the primate retina, with a high resolution region in the fovea surrounded by a low resolution periphery. Furthermore, we find conditions where these emergent properties are amplified or eliminated providing clues to their function.
Brian Cheung, PhD Candidate (Olshausen Lab)
Towards understanding how extrastriate feedback shapes a V1 neuron’s visual response characteristics
It is well understood that feedback connectivity exists from extrastriate cortical visual areas (e.g. V2, V3, and V5/MT) to the primary visual cortex (i.e. V1), but an explanation of the influence of that feedback in shaping responses of V1 neurons is still illusive. A primary difficulty in developing experiments to better understand the role of feedback is a lack of a guiding theory for how feedback could be used in a neural representation of natural visual scenes. In this talk I will discuss a theory of how a hierarchical neural network can use feedback connectivity to improve the network’s representational power. Our work reinforces the assumption that extrastriate feedback is a vital (and non-linear) component of the visually responsive V1 cell’s activity and that the influence of this connectivity is most evident when using natural, ecologically relevant visual stimulus.