CCN Colloquium: "Modeling the neural mechanisms of attention in artificial neural networks"

Countless behavioral studies have demonstrated how validly-cued visual attention can enhance performance on challenging tasks. Neural recordings taken during such conditions of attention have identified the mechanisms that may support this enhanced performance. Given that convolutional neural networks can be thought of as neurally-plausible task-performing models of visual cortex, we can use them to connect neural mechanisms of attention to increases in task performance. I will present a series of completed and ongoing studies that demonstrate this principle. Topics covered will include spatial and feature attention, the relationship between attention and learning, the circuit mechanisms needed for attention, and how these findings can extend to the auditory system as well.