HTC Seminar Series #4: Deep Learning to Solve Challenging Problems
This weeks HTC Seminar Series presets a talk by Jeff Dean from Google presented at Google I/O 19. This talk will highlight some of Google Brain’s research and computer systems with an eye toward how it can be used to solve challenging problems, and will relate them to the National Academy of Engineering's Grand Engineering Challenges for the 21st Century, including the use of machine learning for healthcare, robotics, and engineering the tools of scientific discovery. It will also cover how machine learning is transforming many aspects of our computing hardware and software systems.
Like the Doublemint Twins touting the joys of Doublemint gum, 2 GANs are surely better than 1 GAN. Especially if we package them together inside of one meta GAN module. And this is exactly what the CycleGAN architecture does. Have you ever harbored dark secrets of turning a horse into a zebra? The CycleGAN was developed to do just that. Learn how to turn a horse into a zebra. And more. Now right away you can notice a difference between the image to image transformation GAN architectures we've been discussing over the last few posts. Those last few posts described systems that learn from a database of matched input-output image pairs. And if your goal is to turn an edge representation into a nicely filled in continuous tone image, it's easy to build your database of matched input-output image pairs that your GAN system can then learn off of. Take a continuous tone image (which will be the output of the database pair entry), then run it through an edge detector algorithm.
I thought following up yesterday's TraVelGAN post with a Pix2Pix GAN post would be useful to compare what is going on in the 2 architectures. Two different approaches to the same problem. I stole this Pix2Pix Overview slide below from an excellent deeplearning.ai GAN course (note that they borrowed it from the original paper) because it gives you a good feel for what is going on inside of the Pix2Pix architecture. Note how the Generator part is very much like an auto-encoder architecture, but rebuilt using the U-Net architecture features (based on skip-connections) that fastai has been discussing in their courses for several years before it became more widely known to the deep learning community at large (and which originally came from an obscure medical image segmentation paper) . So the Generator in this Pix2Pix GAN is really pretty sophisticated, consisting of a whole image to image auto-encoder network with U-Net skip connections to generate better image quality at highe
Cairo is a D vector graphics library that is nicely setup for both drawing to the screen or an image as well a generating postscript, PDF or SVG output. We'll be discussing it when we take a look at Cinder as an option for C++ application development in a later post. So i thought it would be cool to build an interface to Cairo for working with 2D graphics in Jupyter notebooks. Which i assumed meant building a Python library interface for it. And low and behold, smart people had already taken on the task. There are several different options to chose from. Let's take a look. Pycairo is the first one to look at. Python bindings for Cairo, just what the doctor ordered. The Pycairo bindings are designed to match the cairo C API as closely as possible. Cairo itself is kind of old school C in it's api design. If you already have Cairo code you want to just use in a Python Jupyter notebook, this may be the way to go. Here's an example of how to use Pycairo. #!/usr/b