Chapter 1 : What is Machine Learning?

“What is Machine Learning” is a question that has been answered in over thousands of blog posts around the globe! However, before diving into learning the technical part of machine learning, it’s crucial that you are familiar with an overall idea of machine learning and its applications.

So what exactly is Machine Learning?

In simple terms, Machine Learning is the science of giving computers the ability to learn, think, and act like a human being! It’s a data science technique that is used to predict behaviors, outcomes, trends by analyzing existing data. Machine Learning allows computers to learn without being explicitly programmed.

Where can Machine Learning be found?

The real question is, ‘where can’t we find machine learning?’ and the answer to that is, NOWHERE! Today, Machine Learning is found in almost every sector there is. Machine Learning Engineers have been able to integrate Machine Learning into all sectors from the Financial sector to the Healthcare sectors!

Why is Machine Learning important?

Imagine you were given a bunch of images containing pictures of people. Now imagine I tell you to write code to identify if a particular person is in those photos. I’m sure if you are a programmer you’ll know that this is an extremely hard task to do! However, if you learn machine learning this task becomes much easier to do.

Every second that passes by, the world creates gigantic amounts of data, and machine learning is an extremely powerful tool to analyze these data and help generate insights, obtain predictions, etc… For example, all your song and movie recommendations on websites like Youtube and Netflix are done thanks to Machine Learning.
Now that we have a rough idea on what machine learning is, let’s get a bit technical, shall we?

How does Machine Learning work?

Machine Learning can be classified under two main catergories:

• Supervised Learning
• Unsupervised Learning

Supervised Learning

As the name suggests, this category implies machine learning under supervision. Here we provide both the input and desired output using classes or tags (Training examples). Then the system infers a function using the provided set of training examples which is used to later predict the correct response when new examples are supplied.

Supervised Learning is further divided into Regression problems and Classification problems.

Classification Problems : The provided inputs are divided into two or more classes, and the system must produce a model that assigns completely new inputs to one or more of the above-declared classes.

Regression Problems : A supervised learning technique where the outputs are continuous rather than discrete.

Sounds like jargon? Let me break it down for you.

Let’s start with Classification! Consider this simple example- Hypothetically assume that a tumor being malignant depends on the tumor size and the age of a person. Now consider the data set shown below :

From the above data we can obtain the following graph:

Here we have represented the benign tumors in blue and the malignant tumors in yellow. Machine Learning is used to self adjust a function to fit the above data as shown below :

Now when we provide this trained model with new input, it will be able to determine if the tumor is malignant or benign. That is, (in this example) if the input data is represented on to the left of the function line, it can be categorized as benign and if it is represented on to the right side of the function line, it can be categorized as malignant.

And that is classification! Now let’s move on to regression; Suppose you are given a bunch of pictures with corresponding ages. You can use a ML algorithm to train a model according to that data. Thereby, when you provide a new picture the system will be able to predict that person’s age!

Unsupervised Learning

Unsupervised Learning basically finds a hidden pattern or intrinsic structures in data. This category does not use labeled data and the most commonly used unsupervised learning technique is Clustering.

If you use google photos, clustering is not hard to imagine. Have you noticed that all your photos are categorized into folders with each folder representing a particular person? So when you click on one person you can get all the pictures in which that person is present.

That’s clustering! It clusters data after finding a hidden pattern in them. So basically unsupervised learning makes the machine figure out all the hidden relations, features, and behavior.

If you have come this far, Congratulations!
You now know the fundamentals needed to get started on your machine learning journey! In the next chapter, we will be exploring data preprocessing in both R and python in order to get started.