In this presentation, I talk about my experience creating the Pknn.jl library for modelling probabilistic K-nearest neighbours and how Julia solved running a sampling an algorithm in Python for hours to only a couple of minutes.
Probabilistic K-nearest neighbours were first proposed by Holmes and Adams in 2001. Their model assumes a distribution on the number of neighbours required to perform inference over a datapoint.
In this presentation, I talk about the practical aspects of the model, and the effortlessly transition from Python code to Julia code for both speed and parallelism.
The project can be found at this link.