News for hackers. Stuff that matters.
Re:Show HN: Scottish Independence Referendum Heatmap by tlarkworthy on Monday August 25, 2014 @08:40AM (#8221494)
The kernel size (smoothing width) should be chosen based on something like leave-on-out cross validation. At the moment the sampling is too sparse and everything is independent and not smooth at all.

This is described in the intro of "Assessment of probability density estimation methods: Parzen window and Finite Gaussian Mixtures"

http://www0.cs.ucl.ac.uk/staff/c.archambeau/publ/iscas_aa06....

(as well as many other places that just happened to be the first google result)

email me if you want more details, I am ex-informatics.

Re:Show HN: Scottish Independence Referendum Heatmap by chriswait on Monday August 25, 2014 @09:28AM (#8221627)
This is massively useful, I was looking for a way to 'aggregate' results somewhat. I will read through this today and might send you a question or two later on if you don't mind.
Re:Show HN: Scottish Independence Referendum Heatmap by tlarkworthy on Wednesday August 27, 2014 @10:57AM (#8232438)
Yeah send any queries over or I can go through it face to face if you want.