We’ve just posted a brief new arXiv article (https://arxiv.org/abs/1604.07078) on learning to represent modulated radio signals using unsupervised learning. We employ a small autoencoder network with convolutional and fully connected layers to fit a sparse signal representation with no expert knowledge or supervision. Mean squared error reconstruction distance and regularization are used during training.
One example of a noisy test set example, its compressed representation, and its reconstruction is shown below for a QPSK signal, additional details are available in the arXiv paper! We achieve a 16x compression in information density (2x88x4->1×44), and 128x in storage space (2x88x32->1×44)! We’re looking forward to doing many more things with these ideas!
As a side note, since drawing hundreds of neural network connection lines in diagramming tools manually is really not fun, I’ve posted a small tool called NNPlot on github which attempts to make generating high level conceptual neural network diagrams much easier. Hopefully someone else will find this of use some day, the network diagram above is the first example in it.