Reverse-engineering neocortical intelligence
Within the MICrONs project, we are using detailed measurements of function and structure of mouse visual cortex to reverse engineer the inference and learning algorithms of the brain. We have recorded optically from 10^5 neurons (thousands at a time) of a behaving mouse across all layers of visual cortex using 2- and 3-photon microscopy. We then use EM to reconstruct the nanoscale wiring of this circuit, and synthesize these diverse measurements in the context of probabilistic inference to relate distributed computations to algorithms. Then we will apply these new algorithms to real-world computer vision problems. Particularly, I am interested using recurrent convolutional networks as a system to test these new algorithms and try to understand its ability to modify either the training or inference in video tasks.