RECENT POSTS RSS

 

Your destiny is written in the stats

Posted 22 July 2013 by Nicolau Werneck

This is yet another article talking about the binomial distribution and statistical hypothesis testing... I read this very silly article the other day, saying that a police department from some Canadian city calculated how many people from each zodiacal sign they arrested in 2011. The article concludes people from the sign of Aries are more likely to be arrested. This is all, of course, just a load of crap. But I was happy to see that the article had a... Read more

Unconfident Bernoulli

Posted 20 July 2013 by Nicolau Werneck

I haven't written for many months... And coincidence or not, when I come back, I have a very similar topic from my last post! This article talks about the uncertainty in the estimation of the probability p from a Bernoulli distribution. In other words, given x successes from N trials, we know a good estimate for p is p~x/N. But how sure can we be about this number? How high or low could this parameter possibly be to give you... Read more

Bernoulli Freak

Posted 1 November 2012 by Nicolau Werneck

I have been reading about hypothesis testing. So when Shirley Manson tweeted this: Sophie Muller just pointed out that on our last two albums, 8 song titles have begun with a B. Itunes playlists can be instructive. Sx I immediately got myself thinking: Is it really significant? Or is it one of those things that are actually easy to happen, like having two people whose birthday is in the same day in a classroom? So I decided to make the... Read more

Metropolis algorithm demo

Posted 28 October 2012 by Nicolau Werneck

This post is about the Metropolis algorithm, a Markov-chain Monte Carlo (MCMC) method to produce samples from a given distribution. This is a very powerful method, but using the algorithm requires some tuning of parameters, and also dealing with compromises to archive the best results. I made an experiment that demonstrates that. My main reference is this MCMC primer by B. Walsh. After reading that I wanted to see how the results of this experiments would be, so here it... Read more

Experiments in sorting algorithms

Posted 30 September 2012 by Nicolau Werneck

This article is about sorting algorithms, and it includes some performance tests for Quicksort I did by myself. Sorting is a big deal in Computer Science. First of all because this is a very important problem. We sort thing in out life very frequently, specially to organize sets of things like product files, student tests or a deck of playing cards for easy retrieval. Sorting is also important in some numerical applications such as finding the approximate median of a... Read more

Neil Armstrong, a manual hero

Posted 27 August 2012 by Nicolau Werneck

Sad news for mankind as one giant hero passed away last week. I must confess I wasn't fully aware Neil Armstrong was still alive. So for me it was one of those strange situations, we first get happy to find out that person was actually alive, and then you get sad with the news, and even more sad to think you didn't know the person was alive. But can you blame me? He was quite reserved. I knew for sure... Read more

Stereographic projection awareness campaign

Posted 19 August 2012 by Nicolau Werneck

I have been working with different kinds of image projections lately, from the most usual like the perspective projection, to often strange-looking projections like the equirectangular. I have found that there are many cases where we need a wide-angle projection where the best choice is the stereographic projection. It is a relatively less known projection, so I am campaigning for the more widespread adoption of it! ... Read more

Facing your deadline

Posted 1 August 2012 by Nicolau Werneck

Just as we switched from the old Nature blogs platform to SciLogs, another change happened in my life. I finally finished writing my thesis! ... Read more

Bicubic interpolation demo

Posted 6 December 2011 by Nicolau Werneck

This is a reproduction of a couple of graphics from an old classic article about cubic interpolation for image processing - Cubic convolution interpolation for digital image processing by R Keys (1981). The version available on-line is a bad scan of what looks like is already an originally a bad printing, with heavy dithering. I am going to use this kind of interpolation here in my work, and because this is a very important topic, and those figures are very... Read more

The parabola behind the exponential

Posted 26 November 2011 by Nicolau Werneck

When you start working with statistical inference and probabilistic models you soon learn that the Gaussian distribution (aka “normal” distribution) is used for a lot of things. Some people get scared by its seemingly complicated formula, but behind that exponential surrounded by fractions there are many simple concepts. This article is an attempt to illustrate that, and also a warning, and example of a case where you might prefer to pick another model for your problem. ... Read more