Naive Classification

I was recently asked about Diogenes’s declaration and about the absolution of classification; specifically, I was asked whether there is an absolute number of attributes that would facilitate proper classification. What follows is my mental model of classification systems.

Classification is the act of relating two or more items to one or more groups. If we only need to form a single group, then zero classifiers are required—everything simply belongs to that group. There is a lot of interesting thinking around implied or enforced groups, from which one must infer classification, but I think this is the opposite of how we typically classify: taking an item and placing it into one of the many groups that already exist in our mind.

Once we embark on the path of differentiation or distinction, the number of classifiers required is implied by the number of groups we intend to create, the attributes available in our collection, and the amount of information each attribute carries. The classifications become combinatorial—we need as many combinations of attributes as we have groups. Binary math makes this easy to visualize, and dichotomous classification is common: with two binary attributes, there are potentially four groups, and so on.

Orthographic projection allows us to classify using a smaller subset of the available attributes, reducing the dimensionality of the dataset.

• Previous: Just Do It and Other Bad Advice