Well folks, there you have it! In this chapter, you learned about RBMs, a little bit of graph theory, and how to create and train a Deep Belief Network in C#. Your buzzword-compliant checklist is almost complete! I recommend that you experiment with the code, train your network layers to different thresholds, and watch how your computer dreams while reconstructing. Remember, the more you train the better, so spend time with each layer to ensure it has enough data to do an accurate job of reconstruction.

A quick note of warning: if you enable drawing of your feature detectors and reconstructed inputs, you will notice a huge decrease in performance. If you are trying to train your layers, you may wish to train them without visualizations first in order to reduce the time required. Trust me, with visualizations it will feel like an eternity if you train each level to a high iteration! Feel free to continually save your network as you progress. Good luck, and happy dreaming!

In the next chapter, we will learn about micro benchmarking and get to use one of the most powerful open source micro benchmarking toolkit ever written!