What is Machine Learning? (Explained Simply)

 What is Machine Learning

Every day you use machine learning. You are practicing machine learning every day.

If there is a show on Netflix list of things they think you will like or when Gmail keeps the spam emails out or when you unlock your smartphone with your face you use face unlocking on your smartphone.. When Spotify suggests a song that you really like that is machine learning too.

What if someone asked you "What is machine learning?" Now how would you answer that in one sentence? Most people do not know what to say because machine learning seems like a thing to learn.

However here is a secret that not many people know: machine learning is not that hard to understand. You will see that machine learning is actually pretty simple.

Over the 6 minutes you will learn about machine learning. By the end you will understand machine learning completely. We will explain the ideas of machine learning and show you how machine learning works and why machine learning is important, for your job or your business or your life.

Ready? Let us start learning about machine learning.

The Simple Definition: What is Machine Learning

Machine Learning: Learning from data, not code.

Let's explain what that means. Normally, if you want a computer to do something you: 

Something, you code what it's supposed to do. Step 1. Step 2. Step 3. The computer can take your directions to the hilt.

It's a different situation with machine learning. Rather than how to write instructions, 

you give the computer examples of data, and it figures out the 

pattern itself.

For instance: A child has to know how to identify dogs. You don't give them 

A 'rulebook' containing 'dogs have 4 legs and a pointy ear and bark. 

You show them lots of pictures of dogs (and non-dogs) and their brain 

learns the pattern. That's machine learning.

Netflix does also do this. Watching films and TV episodes. Netflix notices 

Which is the Kick Off size you prefer? Based on the basis of millions of others: 

It predicts next that you will enjoy, on patterns basis.

That's the beauty of machine learning: It doesn't need any programmer to add each rule. 

the computer program is learning the rules based on data.

Let's now narrow it down more exactly how it happens.

How Does Machine Learning Actually Work

Learning takes place in 5 steps, and machine learning is one of them. Let's take a look at these together.

STEP 1: COLLECT DATA

First of all, you need data. Lots of it. If you wish to foresee whether an e-mail will be read, 

If you are looking to distinguish spam from, you first have to find thousands of spam emails and real emails. 

To forecast house prices, data on thousands of is needed. 

houses: size, location, how old is the house, how many bedrooms, etc.

STEP 2: PREPARE THE DATA

Raw data is messy. Contains mistakes, repetition, lack of content. 

It cannot learn from it until cleared up. 

This usually takes up 80% of the job (and it's dull).

Proceed to step 3 by selecting an algorithm.

An algorithm is a math formula. There are hundreds of them! Different 

Each algorithm is preferable for dealing with a different type of problem. It is as though you are picking the 

right tool for the job.

STEP 4: Train the model

When the magic happens, it's now. The algorithm gets the data fed into it. It studies 

the patterns. It fine-tunes itself for individual use (so-called "weights") 

Loops will be more reliable in predicting the pattern. With that, it repeats itself thousands of times until 

It's an excellent pattern recognizer.

STEP 5: TEST & USE

Lastly, it is tested against new data which it has never seen before. If it works well, 

you use it to make predictions on.

REAL EXAMPLE: SPAM FILTER

Gmail's spam filter works like this:

2. Gmail compiled a lot of spam + not spam emails (millions).

2. They scrubbed the data

3. They selected their algorithm of text classification

4. They took it to those millions of emails

5. Now when you get a new email, it predicts: spam or not spam?

Machine learning is just that. Pattern → Learn → Predict.

Then, we will discuss various types of machine learning.

Types of Machine Learning

We have three types of machine learning.

SUPERVISED LEARNING

You provide a set of examples with labels to the algorithm. It's a known solution.

It learns how to guess the answer to an unknown set of information not yet labelled.

The problem: Forecast the average selling price of a house. You present the algorithm 10,000 

The combination of houses with their characteristics (house size, location, age) and their clientele (real). 

selling price. It catches on. Now, when you are providing it a new, 

Predicts the price of the house inside the house.

Other examples: Email spam detection, disease diagnosis, predicting 

If a customer is going to go out, he or she will.

UNSUPERVISED LEARNING

You provide algorithm with data UNLABELED. It has to find patterns 

on its own. Finds secret groups or structures.

Example: Customer segmentation. You provide Netflix with information on what. 

It was seen by (no numbers). Netflix groups customers 

Automatically classified into categories: Action lovers, Comedy watchers, 

Documentary binge-watchers or otherwise, never tell Netflix. 

It found these groups; what they were.

Other examples: Clustering social media users, identifying market. 

segments, fraud detection.

REINFORCEMENT LEARNING

It is a learning algorithm, that is, it learns through trial and error. As it is rewarded for good 

made decisions, and was penalized for wrong ones. Train a dog with rewards!

Example: Google's AI, AlphaGo, learning to play the game Go. Every winning move 

gets a reward. Loser takes a penalty for all of their losing moves. Those are just a small few of millions of games, 

it's improved over the best human player in the world.

Self-driving cars, game playing AIs, robot learning are additional examples.

COMPARISON

In most real world applications, supervised learning (answers) are used. 

The great thing about unsupervised learning is that it is useful for learning patterns! Reinforcement 

Learning is newer and might be stronger but more difficult to apply.

Let's now consider some more real life examples you may have heard before.

Real-World Machine Learning Applications

Here are some applications you would come across which you use every day (probably without ever knowing it).

HEALTHCARE

Predicting diseases at an early stage by doctors with the help of ML. Multillions of examples. 

This new technology of chest X-rays can now detect lung cancer earlier than some human. 

radiologists. This saves lives.

FINANCE

Using ML, banks can identify fraud. Algorithms help you determine the price of whatever you're buying. 

instantly decide: Is this a fraudulent transaction or legitimate? 

It measures and processes hundreds of factors in millisecond time.

E-COMMERCE

Amazon makes use of ML to suggest products. When you are surfing, algorithms 

observe your behavior, make comparisons with similar customers and make predictions 

what you'll buy. Amazon's recommender system is making billions of dollars for Amazon.

SOCIAL MEDIA

Instagram's News Feed is built with ML. What posts appear? It's not chronological. 

A system of rules that tells a computer how to work out answers to problems. A set of instructions used by a computer to calculate solutions to problems. 

This way you won't stop scrolling (or screen).

TRANSPORTATION

In fact, the autopilot at Tesla is powered by ML. Cameras capture video. Algorithms instantly 

recognize : Is it a STOP sign? A pedestrian? A car? This happens in 

To pass 65 mph with a speed of just milliseconds.

Machine learning is NOT science fiction; it's your everyday reality now.

So here's the question: Is machine learning right for you?

Know the Difference Between AI vs ML vs DL : READ MORE