One typical task carried out in the field of Data Science is classification. This means using response attributes to classify (tag) different objects or individuals. The name of our new company, "One on Epsilon" does just that. Let me explain.

Whenever I tell someone about the name One on Epsilon, I get one of two typical responses:

Response 1: An understanding smile followed by a statement like, "Geez, that's a huge number", or "what a cool name!".

Response 2: The person replies with something like, "One on what?".

The name of the company thus serves as a basic classifier differentiating between those who are used to think of variables denoted by the Greek letter, Epsilon, as representing extremely small positive real numbers; and those who simply have not touched this type of thing previously. That is, people with a response of type 1 are typically engineers, scientists, mathematicians, statisticians or are generally well versed in mathematics. As opposed to that, people responding with response 2, are probably not. And that is certainly fine. For this latter group, let me explain a bit about epsilon in mathematics.

Epsilon is typically taken in mathematics as a very small quantity. In mathematical analysis, it is actually taken as an arbitrarily small quantity. This allows to formally define notions of convergence as used in Calculus, Analysis and related fields. To say that a sequence of numbers converges to some limit, one would then say that for any arbitrary small epsilon (as close to zero as we would like), there is some point in the sequence, from which on wards the distance between each element of the sequence and the limit is less than epsilon. This type of thinking was formalized by many great mathematicians during the 19'th century. It serves as the basic concept for dealing with infinity as well as infinitesimally small quantities.

Alternatively, small values of epsilon are often prescribed in applied contexts. When running computational algorithms in Data Science, optimization or many other fields of numerical mathematics, convergence of the algorithm to the correct solution, is often specified by fixing a parameter named Epsilon at some predefined fixed value. For example epsilon = 0.000000001. Then some algorithm is run, until the changes in the variables of the algorithm are smaller than epsilon, at which point the algorithm stops and yields a result.

So if epsilon = 0.000000001, what is 1/epsilon? It is a Billion. That is quite big!