Blog

Mathematics of Convolutions

Convolutions are a mathematical tool of translation invariant transformations to objects. It’s best to think about 1 dimensional convolutions before thinking about convolutions of 2 and 3 dimensions. A convolution of two functions \(x[n]\) and \(h[n]\) produces a new function \(y[n] = (x\star h)[n]\). The operation \(y[n] = \sum_{k=0}^{N-1}x[k]h[n-k]\) where the \(x\) has support over […]

Dilated Convolutions

In this post, I wanted to try a new type of problem. This is the problem of semantic segmentation. Semantic segmentation is essentially a method of assigning categories to different parts of an image. The best way to explain it is through a picture: There are several methods to find accurate segmentations of an image. There […]

Manifold Learning

In this blog post, we will apply the methods of nonlinear dimensionality reduction (also known as manifold learning) to the Kepler dataset, using the paradigm of avalanches as described in the first post. The entire set of code is on github, please check it out! We have about \(16,000\) stars to process, with each star containing a […]