Introduction to Machine Learning

Team PolyrificApril 10, 20176 min read

Introduction to Machine Learning

Machine learning is a system of algorithms aimed at detecting patterns in big data and then learning from those patterns without being explicitly programmed by a human operator. These algorithms take a probabilistic, rather than a deterministic, approach to accomplishing goals. Let's take a quick look at what that means by way of example:

Deterministic Approach

A human programs software in no uncertain terms to remind him to bring his umbrella to work if the chance of precipitation in his area is greater than 40%:

//Get the weather forecast
var chanceOfRain = myWeatherForecast().precipitationChance;

//Send the message
if(chanceOfRain > 40%){
    sendEmail("Bring your umbrella to work!");

This approach is deterministic because the outcome is predetermined by the author of the code in no uncertain terms.

Probabilistic (Machine Learning) Approach

An algorithm crawls big datasets such as tweets about the weather, weather news, forecasts, umbrella sales, and supervised feedback (we'll get to that later) such as, "did you bring an umbrella to work today?":

machine learning

As you can now see, Machine Learning is about statistics and probability. Whereas humans typically write algorithms with a predetermined output for a given set of a few inputs, machine learning algorithms consume a vast amount of inputs and surface the most likely output. Over time, machine learning algorithms get better by adjusting their own internal methods to predict outcomes based on a growing body of data or knowledge. Eventually, a machine learning algorithm can become better at predicting outcomes than can its human counterparts because human beings, and deterministic algorithms written by humans, can not factor in the massive quantities of inputs that machine learning algorithms easily digest.

Machine learning is nothing new

Believe it or not, the early concepts of machine learning and statistical analysis were proposed by British mathematician Alan Turing more than sixty years ago. So why is machine learning just now becoming so popular? There are several factors for this but chiefly it is because processors are more powerful, data storage is incredibly cheap, and thanks to the Internet, everything is now connected. In other words, it is now common for the phone in your pocket to display to you a message such as "bring your umbrella" as it detects you are leaving your home for the day as a result of a server to which it is connected crunching potentially millions of connected data points. Even more impressive, this happens in less than a second and it is all thanks to machine learning.

You use it every day

You might not have realized it, but you already benefit from machine learning every day. Here are a few of the common areas in which machine learning is making our daily lives better:

Search Results

Think that Google and Bing get your search results simply by how many times your search term is displayed on a given web page? Think again. Such a deterministic approach would never help you wade through the universe of irrelevant information to get to what you actually need. Search engines use complicated machine learning algorithms that factor in information such as your current location, the time of day, your recent searches, the page you last visited, your recent purchases, and so on. This probabilistic approach usually surfaces what you were probably after within the top ten results on your page. A deterministic approach would never succeed in doing the same.

Spam Filters

Every time you classify an email as "spam" you are training a machine-learning algorithm to recognize similar emails as having a high degree of probability of being spam.

Recommendation Engines

Netflix, Amazon, and even Spotify use your searches and feedback as fodder for machine learning engines that in turn offer you better recommendations.

News Feeds

You may thank that your Facebook feed is already full of posts that don't hold your interest, but this would be even worse without machine learning which is at least promoting to the top of your feed people and topics you seem to care about the most.


You may have already heard of Google Translate which helps you translate text from one language to another but it gets even better: Microsoft recently capitalized on machine learning algorithms to translate spoken Skype conversations in real-time. This means you can literally talk to someone in English while they hear Mandarin and vice versa.

Fraud Detection

Ever had your credit card declined when you are traveling or making a large purchase? Machine learning algorithms help credit card companies detect events that seem out of character for you and decline the card to prevent fraudulent use.

Will machine learning replace humans in the workplace?

Probably, but it should be considered as good news because it will promote productivity and creativity. In some cases, machine learning will augment workforces wherein there are currently shortages of skilled workers. 

Consider the example of a general practice physician. Currently, you will need to schedule an appointment with your GP when you have a medical issue and, once you get in to see the doctor, you have to carefully explain your symptoms and answer the doctor's questions. Drawing on her previous experience the doctor may be able to make a diagnosis, but if she happens to not have prior experience or knowledge about your particular situation, then she may refer you to a specialist or send you to a lab for tests. 

Now consider a digital doctor which is trained by vast quantities of shared data and experience. At any given time, this digital doctor may be processing--and learning from--millions of cases. This doctor becomes more experienced and better at making correct diagnoses by the second. So when this doctor interviews you and makes a diagnosis, it is drawing on millions upon millions of past cases, symptoms, and outcomes in order to let you know what is ailing you in a matter of seconds. There is not likely a need to go for further tests or a need to go see a specialist.

Human doctors can use such a digital doctor to augment, rather than replace, their care of you. Now they can receive a report and know immediately whether your appointment should be expedited, or whether they can simply write a prescription for you to come to pick up.

Get started!

Machine learning is a productivity multiplier. You can think of it as a way to unload cognitive tasks to the computer so you can free your staff's minds to focus on higher-value work. Machine learning is no longer the property of big tech companies alone, we can all now capitalize on the amazing things it can accomplish. Contact us today to learn more about how machine learning can benefit your business.

Top posts
  • The Web is Not Enough
    Introduction to Machine Learning
  • The Basics of Business Intelligence
    Introduction to Machine Learning
  • SDLC Series: Continuous Integration
    Introduction to Machine Learning
  • SDLC Series: NoOps Primer
    Introduction to Machine Learning
  • CES 2018 -- Trends to Watch
    Introduction to Machine Learning