Question: What is the opposite of a fart?
Short answer: A car alarm.
Long answer: Last week we described an algorithm for deciding whether an arbitrary sound clip does or does not contain a fart. Because the algorithm makes use of an artificial neural network (ANN), we call the algorithm FartNet.
FartNet takes a sound clip as input and returns the fart probability as the output. We initially used the algorithm to identify non-fart sounds that resembled farts, but a reader has commented that we could also use it to identify farts that are more fart-like than others. In other words, we could use the algorithm to identify canonical fart sounds. This is something we tried to do previously with principal components analysis.
As requested, we have performed the analysis, and without further ado, here are the three fartiest farts:
The first is from Subject S, a longtime member of the College, who has occasionally contributed to our research. The other two farts were produced by ChillAlien, who is also an active contributor. The farts were otherwise unremarkable, having VOC levels of 0.2, 7.5 and 3.5 ppm, respectively. But all three farts were given a rating of greater than 96% by FartNet, making them the fartiest sounds in our database.
These farts are therefore representatives of all other farts, and in that sense they are similar to what the philosopher Plato had in mind when he wrote about pure forms.
For completeness, we have also used FartNet to determine the non-fartiest non-fart. This is the sound that the algorithm determined was least likely to be a fart. The answer turned out to be a car alarm, which was estimated to have a 4.5% chance of being a fart.