The forward-forward algorithm looked interesting to me. This page is going to be a summary of my notes about it if I ever get the chance to play with it.
You can find the paper by Geoffrey Hinton at https://www.cs.toronto.edu/~hinton/FFA13.pdf.
Instead of training a whole deep neural network all at once with backpropagation, the forward-forward algorithm trains each layer of the network one at a time, in a greedy fashion.