I have been trying to wrap my head around the way PixelCNN++ is using discretized mixture of logistic distributions to sample pixels and could not find a single comprehensive source about it, so I wrote a blog post incorporating everything I learned along the way
Month: May 2020
Learning Shape Abstraction by Assembling Volumetric Primitives
One of the best ways to know that you understand a subject is when you are able to summarize it in a concise manner. Similarly, one of the recurring themes in scientific research is to explain a phenomena using simple patterns. These simple patterns allows us to interpret, predict and manipulate the world around us much more easily than if we chose more complex representations.