They threw 10 million random images (fr
They threw 10 million random images (frames of YouTube) at a neural network and found that the neurons had self-organized themselves to put the images into different categories. One neuron was especially good at recognizing human faces, another at human bodies, etc. Amazing implications.
Building high-level features using large scale unsupervised learning
Shared with: Public
Reshared by: TheBlack Box
This post was originally on Google+