Some Miscellaneous Stand-Alone Machine Learning Codes. Mostly from TensorFlow and SK-Learn.
Some codes here may not work as stand-alones, like AI_music1, but helpful chunks to get started with. (AI_music1 may actually work for you and may just be my depencies acting wonky when I tried it in terminal)
Some codes written in Google's jupyter-style Collaboratory framework. Nice layout for writing and testing ML codes, but seems a bit slower than Kaggle's similar 'Kernels' framework, have to test it out more. Apparently they're letting users access Google's TPU's through collaboratory with a streamlined process for dealing with the TPU exclusive to the framework. Post on this more as I test it out, may put TPU computing in its own repository.
Seems to have issues with libs not common in numerics, like those for managing audio files.
Also test out and present examples of Mathematica's machine learning fraamework (post as pdf with code included). Pre-trained MNIST Neural Network built-in, so its nice and easy to use, but I suspect that actually training NN's on Mathematica would be horridly slow given Mathematica's architecture.
Does, however, present some nice options for visualization. The NN visualization techniques built into the ML packages are very lackluster but Mathematica does otherwise have some very nice and intuitive tensor/matrix visualization techniques. Visualization on Mathematica a lot easier than standard libs, including MatPlot (but typically not as high-quality), so mathematica may very well be a good niche tool for prototyping visualizations of NN's. More to come on this.