##Documentation
Simple code to demonstrate Direct Feedback Alignment training of neural networks. Network with 2 hidden layers is trained to perform the xor computation. Same weight initializations trained using both DFA and Backpropagation for comparison.
DFA algorithm is implemented as described in this paper.
dfa.m will run and plot all results
##Figures
Sample training
Average Convergence
##References
Nøkland, Arild. "Direct Feedback Alignment Provides Learning in Deep Neural Networks." arXiv preprint arXiv:1609.01596 (2016).