From time to time I write slides on some subject, to present something I’ve studied or for teaching. Last slides have been to introduce Tensorflow in a graduate course, in form of Python notebooks and Html. They cover

  • basic usage examples
  • typical layers
  • variables and how to save and restore a model
  • feeding placeholders and using queues to speed it up
  • selecting one GPU if you have more
  • more interestingly, walkthroughs on code implementing fine tuning a VGG, VAE, GAN and style transfer

A second batch of slides is on learning graphical models. They cover probabilistic learning (taken from Nowozin and Lampert tutorial) and an exercise of CRF learning with Structured SVMs.

Finally, I spent some time reading the Rasmussen and Williams book on Gaussian processes plus some papers using them, and compiled what I learned in these slides.