A very prominent saying that I truly believe in is, “The more you learn, the more you earn”. It applies everywhere in every aspect of life. So be it in learning anything of/about machine learning.
In this article, I will share a different perspective of how you can apply the tactics to learn any concept of machine learning. Will also be showing it with an example of a simple algorithm.
Let us take a look at a few stats of learning and also the best way to learn.
We learn 10 percent of what we read, 20 percent of what we listen to, 30 percent of what we see, 50 percent of what we see and hear 70 percent of what we discuss with others 80 percent of what we personally experience 95 percent of what we teach others.
So the best way to learn any concept is understanding it yourself and then teaching it to someone else. With this you will be able to soak the most of the content that you have learned!
Now let us see how I would be doing it if I didn’t know anything about Support Vector Machine(SVM).
Note: Here even if you know the basics of SVM, it is still advised to read it up whole so as to give you a brief understanding of how you should reach a problem/concept in machine learning.
Understanding it better by revising SVM
So if I were to learn about SVM, I would proceed something like this.
Step 1: Basics are important
First I would start from its full form, that is support vector machine. Then would try to understand why it is called so? Well, I would keep that up for the last because to understand that I would be in a need to understand the intuition behind the same first.
I would read from the basics and understand that this algorithm supports both linear and non-linear regression. Also can be used for classification purposes.
The main aim of this is to get as many data points as possible while limiting the margin violations.
Step 2: Learn by asking questions
Along with this, I would learn what is the difference between a separable class and inseparable class. Like in the image above we are able to understand that it is an easily separable case but what is the case is not linearly separable. Would study about how kernel tricks come into the picture?
Step 3: Learn by tweaking parameters
In SVM for classification, I would look at the terms that are used the most and would also understand the tuning parameters, like margin, kernel, regularisation parameter(c), gamma etc.
The example is shown in the image below.
Some images of different kernel parameters.
Source: Scikit learn
Step 4: Learn by implementing
Now when you have got a better understanding of the topic, learn by implementing whatever you have learned. By doing this you will learn from the mistakes you make. You can implement whatever you have learned by solving questions on classification where you can take the dataset from Kaggle or google datasets. With this, you will not only practice classification/regression-based problems but also would get a better idea of how to deal with different types of data.
Ask as many questions as possible whilst learning the intuition, because once you get into the sklearn’s inbuilt functions there’s hardly any growth there.
There are a lot of different points of SVM which were left untouched because this article is dedicated to telling you the way you have thought. The main plot is to benefit the readers with a new strategy than explaining them about SVM.
The crux of the article
Always ask the following question to yourself whilst learning anything.
- Why is it important to learn?
- Why is the name of the concept so?
- What is it?
- Where is it used?
- How can I tune/engineer it?
- Finally this quote: “If you can’t explain it to a six-year-old, you don’t understand it yourself.”
― Albert Einstein
This article was aimed to give you a brief idea on how to learn any concept in machine learning. It is taken into consideration to show it with an example which would help the readers understand better on how or should be using the same tactics which can benefit them the most. With this article, you have saved a lot of time as you have got a better understanding of how to ask questions to yourself whilst learning any new thing and also have got a revision of SVM both at the same time.
Join Our Telegram Group. Be part of an engaging online community. Join Here.
Subscribe to our NewsletterGet the latest updates and relevant offers by sharing your email.
Understanding and building fathomable approaches to problem statements is what I like the most. I love talking about conversations whose main plot is machine learning, computer vision, deep learning, data analysis and visualization. Apart from them, my interest also lies in listening to business podcasts, use cases and reading self help books.