2731: K-Means Clustering

2023-01-30T18:36:46Z

Benjaminhwilliams: Use em dashes for subordinate clause; use singular 'criterion' in place of plural 'criteria'.

{{comic
| number = 2731
| date = January 30, 2023
| title = K-Means Clustering
| image = k_means_clustering_2x.png
| imagesize = 320x385px
| noexpand = true
| titletext = According to my especially unsupervised K-means clustering algorithm, there are currently about 8 billion types of people in the world.
}}

==Explanation==
{{incomplete|Created by EITHER 8 BILLION OR 3 TYPES OF BOTS - Please change this comment when editing this page. Do NOT delete this tag too soon.}}

{{w|K-means_clustering|''k''-means clustering}} is a method of categorising ''n'' vectors into ''k'' clusters. For example, we might categorise a population by two metrics, and want to best categorise this scatter graph into the distinct populations, algorithmically drawing {{w|Voronoi cell}}s to decide the within-cluster variances.

Ponytail's determination that there are three clusters is unsurprising if she herself falls into the category of those who use K=3 as a fixed value, which will inevitably result in three data clusters regardless of actual distribution. The qualitative interpretation of the other two categories — that is, what placement in the other two categories means — is unclear as Ponytail's analysis is either using a binary criterion (whether or not one sorts data into three groups) as the basis for sorting people into three categories, or is a black box using unknown criteria and she has only been able to determine that her own group shares the tendency to group things into threes.

The title text refers to a K-means algorithm with the opposite problem, with no reduction of K value to converge any two human beings into a common cluster based on shared traits. This is humorous because it would make such a clustering useless for the purposes for which a K-Means Clustering is typically used, such as of making insurance risk pools or targets of advertisement campaigns.

Interestingly, by including the entire human population, the algorithm should be immune to bias in creating its input data. However, since every human is unique,{{cn}} the only way to have the clusters converge is to "throw out" some traits of humans as unimportant. This may be objectionable to humans who disagree with that assessment. In contrast, in a supervised algorithm, the training data is tagged with traits that the trainers seek. These traits could be applied in a manner that is socially unacceptable, and lead to AI behavior that reflects the biases of the trainers.

==Transcript==
{{incomplete transcript|Do NOT delete this tag too soon.}}
:Ponytail is presenting on a stage, pointing a screen with a stick. The writings and possible figures on the screen are illegible.

:Ponytail: Our analysis shows that there are three kinds of people in the world: Those who use ''k''-means clustering with ''k''=3, and two other types whose qualitative interpretation is unclear.

{{comic discussion}}
[[Category:Comics featuring Ponytail]]

explain xkcd - User contributions [en]

2731: K-Means Clustering