This application of deep learning could help Numenta reach its goals sooner that expected:
https://www.mpg.de/12130750/neural-networks-connectome?c=2249
There is already a machine mass producing brain slice images that I saw in a Simons institute video. With full connectivity maps it should be a lot easier to figure out the algorithms enacted by the brain.
2 Likes
It should be quite a big part of connectome to make it really useful. That’s what we have at he moment as the biggest achievement: https://www.sciencealert.com/scientists-scan-fruit-fly-neurons-for-most-detailed-brain-image-ever
I’m not sure if the insect’s brain has the same motives in neuron organization as the cortex.
That’s pretty cool… using convolution neural networks to segment image slices of the brain to map out neurons.
I know the past few Kaggle competitions have been focusing into image segmentation. Only a matter of time until those techniques, network designs, and weights start getting into newer areas. 
(For those interested, here an open-source network for doing something similar. Just repeatedly do it on a series of images along a specific axis, and you can build up 3d models: https://github.com/matterport/Mask_RCNN)
1 Like