Bootstrapping: four types of it

I am sure you might have heard the term bootstrapping already. The first time I heard it, I felt it like jargon. Later I came to know that there are four different scenarios the term comes in. All of them have some basic roots but are in different subject domains.

The boot-strapping

We all know this. This is the obvious interpretation of the term bootstrapping. We might have struggled in our childhood days to do this when the shoelace gets untie frequently. Some shoelace tieing methods keep the tie intact as long as we want. Among the many different tieing methods, I found this method quick & easy and it just works fine all the time.

Computing bootstrap

Apart from the tieing tip, the process can be used to denote some kind of preparatory step. This is where the idea comes in computing. Anyone from a computer hardware background would know the intricacies involved in starting a computer with loading the Operating System from hard disk to main memory. Wikipedia defines it as a self-starting process with a lot of steps involved in between, like testing hardware, memory, etc., which are all handled by the bootloader program. But I still feel that it takes too long for computers to boot after powering on, especially in urgent situations :-).

Web development bootstrap

bootstrapp logo

The other area where this concept appears is also related to computers. Here developing graphical interfaces for web pages can follow a framework so that the content render good in all screen sizes. Twitter developed this open-source coding framework. As of now, many websites use it to make the content appear responsive to all screen sizes. Maybe the developers used the name to represent the characteristics of the framework that has all the fundamentals qualities needed for a website design.

Statistical bootstrapping

You might be probably looking for this version of bootstrapping. We use sampling in real-life, like asking multiple people about their opinion about a movie, restaurant, political polls, etc. The very first reason for sampling to be needed is that it is impossible to go and ask every person on earth about their opinion on something. It is infeasible to test a drug on everyone so that we are 100% sure it works. However, since the samples could be diverse, by choosing a part of it we might end up with a sampling bias/error even though we did a random sampling.

Bootstrapping deals with this problem. It is a statistical technique where random sampling is done with replacement. The plus point is that it lets us estimate parameters like bias, variance, prediction error, etc. The key idea here is that sampling with replacement somewhat brings the effect of sampling from the population itself. Bootstrap was a landmark development in the field of statistics. It is relevant to us and used many machine learning algorithms as well.


I think there are even more versions of the concept. The applications section of this Wikipedia page has a lot of such use cases. Let me know if you know or encounter some new ones.

Leave a Comment

Your email address will not be published.