Excited to share a new manuscript!

Deep nets are great at predicting visual neurons. Yet, they are unable to tell us which artificial neuron directly corresponds to a biological neuron… until now!

biorxiv.org/content/10.110…

(yes, that is indeed a fictive female fly, good guess!) We used a deep net to model the fruit fly’s visual system, which has a bottleneck of ~40 optic glomeruli (or “channels”).

Each model unit maps to one of the channels---the unit predicts the real neuron’s activity *and* how that neuron drives behavior...a one-to-one mapping.

My new work with Jonathan Pillow @jpillowtime! "High-contrast 'gaudy' images improve the training of deep neural network models of visual cortex."

We found that gaudy images can train DNNs with little data---perfect for neuro experiments!
arxiv.org/abs/2006.11412 #tweeprint Our goal is to predict visual cortical responses from natural images. Often linear regression is used to map image features to responses b/c lack of experimental data. Here, we use a DNN (readout network) and avoid overfitting b/c our gaudy images are tailored for training!