What is Machine Learning?

What is machine learning? What does it mean that the machine learns? Can machines actually learn?

If you are asking these types of questions, then let me help you find the answers. I will take a simple approach to explain what is machine learning. By the end of this article, you will have an overall idea of the concepts behind machine learning. So without wasting any more time, let’s dive in!

Looking for video courses on machine learning? Then check out Applied Machine Learning – Foundations course on Linkedin Learning. Sign up for a free account today to get access.

Machine learning is the field of study or method that allows computers to learn from observational data and make predictions based on it. It gives the computer to learn without the need to be explicitly programmed.

A bit more general definition of machine learning:

“Machine learning is the field of study that gives computers the ability to learn without being explicitly programmed” – Arthur Samuel, 1959.

And for all you engineer out there:

“A computer program is said to learn from experience E with respect to some task T and some performance measure P, if its performance on T, as measured by P, improves with experience E.” – Tom Mitchell, 1997.

Machine learning models learn from past data to make predictions on future data.

An example of machine learning is a spam filter program that learns how to flag spam emails when you feed the system with samples of spam emails and samples of regular email.

Machine learning programs learn to make predictions by finding patterns in vast amounts of data.

What differentiates machine learning program with regular programs is that machine learning programs learn to deal with uncertainties and probabilities. You can tell that the coding process of a machine learning program is completely different than that of coding a regular application. As a developer, you need to be familiar with a new set of concepts and data structures related to machine learning.

AI vs Machine Learning vs Deep Learning

Before we talk about more on what is machine learning let’s draw a line between AI, machine learning & deep learning. Just in case if you may have any confusion.

Artificial Intelligence – AI is the theory of machines performing tasks that require the characteristics of human intelligence. For example, vision, speech recognition, making decisions, etc.
Machine Learning – ML is the subset of AI that allows the machine to learn on their own without being programmed. It’s an application of AI in which the machine learns semi-automatically from data.
Deep Learning – Deep learning is a subset of machine learning based on artificial neural networks capable of learning from data that is unstructured or unlabeled.

The main point to note here is that machine learning & deep learning both fall under artificial intelligence.

Why Use Machine Learning?

For tasks that are too advanced for us to code directly, we can utilize machine learning.

There are tasks that are so complex that it is impractical, if not impossible at some point for us to work it out. So instead, what we can do is feed a large amount of data to a machine learning algorithm. And then let the algorithm explore the data and take care of the problem.

Machine Learning programs or algorithms can analyze large chunks of data and detect specific patterns that would be impossible for humans to detect.

Let’s say you run an e-commerce store online. And you want to understand the browsing behaviors and purchase histories of your website’s users. In order to understand the behaviors of your users, you can implement machine learning algorithms. These algorithms are able to go over huge volumes of user behavior data to help you gain more insights about your users. As a result, you are able to understand your users’ needs and recommend the right products to them.

Some of the other common uses of machine learning, is self-driving cars, credit card fraud detection and even those recommendation engines that you see on Netflix & iTunes.

In summary, machine learning is best for:

Problems require a lot of explicit programming and long lists of rules.
Problems that too complex to solve using a regular approach.
Finding patterns to get insights from large amounts of data.

And of course, there are more. These are just some of the problems for which you can utilize machine learning.

Two Main Types of Machine Learning

You can classify machine learning into two common types depending on the problems you are trying to solve. Each type of machine learning has its own set of algorithms that we can utilize. Here are two common types of machine learning that you should know:

Supervised Learning
Unsupervised Learning

Supervised Learning

In supervised machine learning, you teach the machine by providing training data that has inputs (features) and labeled output (desired answers or results) to make predictions based on future input data. It learns from labeled data from the past to predict a new label, given some features.

It is important to understand that the data that you feed the model must learn from outputs that are labeled. Labeled data means that your training data is tagged with the correct answer. So, whenever you think of supervised machine learning, think of the word “labeled”.

And again, supervised learning algorithms use labeled examples, such as an input where the desired output is known.

Two common supervised learning algorithms are classification & regression. If the labels in your data are continuous then it’s a regression problem. If the labels are categorical then it’s a classification problem.

Classification

Classification problem such as given a person’s height and weight, predict their gender. Here the features are height and weight. And our label is gender, which is categorical. That’s why we are applying a classification algorithm.

If you plot this data then, it will look something similar to this. Remember our label here is male and female, the genders. The height on the x-axis and the weight as the y-axis as features.

After training our model using training data, then we insert a new point into our mode. We already know the weight and height of the new point. However, we don’t know what class of gender it belongs to.

After that our machine learning algorithm predicts according to what it has been trained on, is that the new point falls under the gender male.

Now let’s take a look at regression which is also a supervised learning technique.

Regression

Given the size and number of rooms in a house, predict the selling price. Here the features are size and rooms. We also have our label which is the price. As our label is continuous, we are applying a regression algorithm.

So, if you plot this data, then it should look something similar to this. We are only using one of the features here which is square feet on the x-axis. Our label here is on the y-axis.

One thing to know that about regression problems is that we have our independent variable (squire feet) on the x-axis and the dependent variable (price) on the y-axis. This is obvious as the price depends on the size of the house.

Our model then ends up creating some sort of a line to fit into our data. This helps our model to learn that larger the house, the higher the price.

And now when we want to know the price of a new house for which we don’t know the price, but we do know its feature (square feet) our model is telling us it’s predicted the price.

Just to recap, supervised learning trains the model using historical data where the output or the response is labeled. And once you train your model, you can feed new data where the features are known to make predictions.

Unsupervised Learning

But what if you don’t have historical labels for your data? We only have features.

The answer is, you need to use unsupervised machine learning to train your model.

Unsupervised machine learning algorithms find patterns from a dataset without reference to known or labeled outcomes.

Again, labeled data means that your training data is tagged with the correct answer. Here you have no “right answer” to fit on, you need to look for patterns in the data and find a structure.

Clustering

One of the most common unsupervised learning algorithms is clustering. Clustering is the task of dividing the population or data points into a number of groups such that data points in the same groups are more similar to other data points in the same group and dissimilar to the data points in other groups. It is basically a collection of objects on the basis of similarity and dissimilarity between them.

Clustering problems such as, given heights and weights for different breeds of dogs. There is no label given. Only features. It is up to you to cluster together with the data into similar groups and then interpret the clusters.

If we plot all the features, then we have.

After computing the clustering algorithm, we end up having these two clusters. Our machine learning model groups similar features together between the clusters.

Remember that, clustering cannot give you the group labels.

What it can tell, is that the points in each cluster are similar to each other based on the features.

Learn Artificial Intelligence

Interested in learning more about AI & what is machine learning? Then check out LinkedIn Learning.

If you are an avid learner like me, and constantly looking for ways to improve your skills, then LinkedIn Learning is the platform for you. You can literally find courses on literally “everything”. From coding & machine learning to even on how to make music.

Recommended: Coding for Beginners — Start Here

No need to pay for each course as their monthly subscription model lets you access their whole library. And guess what? After you are able to complete a course, you are awarded a certificate.

How cool is that?!

Listing professional certification on a resume shows evidence that you have the relevant training and expertise to take on the job responsibilities. As a result, it increased the chance of you getting the job.

Read: Five Must-Know Job Interview Tips (Ace Your Next Interview)

The best part is that you can sign up for a 30-day free trial. During this free trial period, you can cancel your subscription any time without being charged. That’s why I highly recommend that you take advantage of this platform.

Sign up today and get free access to their thousands of online courses. Learn about artificial intelligence and how we can utilize it to benefit humanity.

Conclusion

I hope you now have a basic understanding of what is machine learning.

One machine learning type that we did not talk about today is reinforcement learning. That’s a whole concept of its own. However, a good starting point for you is to start with supervised & unsupervised machine learning.

Supervised learning works with data that has a label, while unsupervised learning works with data that does not have a label. And labeled data means that your training data is tagged with the correct answer.

If you are wondering which programming language you should start learning machine learning, then Python is a great language.

Make sure to check out Linkedin Learning, for video courses not only on machine learning but also programming, data science & more!

What comes to your mind when you hear the term “machine learning”? Is it Terminators? Or something else?