What is machine learning? If you’re interested in analytics you might have heard this term a lot. Many misuse it and some predict grandiose outcomes for its future.
Despite the hype, machine learning is one of the most powerful technologies in the modern enterprise. Soon you’ll understand why.
What is machine learning?
Machine learning is a school of computer science that focuses on programming machines to improve their own performance through data and iteration.
In traditional programming, a human engineer needs to code every process the machine performs. This makes solving some types of problems difficult.
Take the example of a video streaming service like Hulu or Netflix. In order to make recommendations for users, engineers need to decide which videos to recommend based on input from users. It’s a case of “if A then B.”
From a certain perspective, this works. The program can pair users to video recommendations based on determined factors like genre. But how can programmers do this for thousands of titles and millions of unique user histories? And can engineers really know someone who enjoys one title will enjoy another? How quickly will the program adjust as viewing habits change?
Machine learning solves these problems. Rather than relying solely on human instruction, machine learning uses algorithms to gather data, learn from them and make predictions. The system can adjust itself as users provide more information.
Often this leads to recommendations a human wouldn’t predict. Maybe fans of a certain horror movie enjoy a particular romantic comedy, for example.
AI vs. machine learning
One of the most common questions people ask is the question of machine learning and AI. The answer is simple. Machine learning is a type of AI. It’s a subset within the larger field in the same way artificial intelligence is a subset within the larger field of computer science.
Unlike machine learning, AI is a rather non-technical term. It’s more about outcome than specific methodology. Any smart system that mimics human behavior is an AI.
Supervised learning vs. unsupervised learning
Supervised learning is a common task in machine learning that works by using input and output pairs to train an algorithm. Let’s use an example of a team that wants to train an application to recognize dog pictures.
Dogs might look very different from one another. It’s extremely difficult to teach a computer to recognize the similarities of a Great Dane standing on a mountain and a Chihuahua swimming in a lake. Any traditional rules you might create for the system would inevitably have exceptions. That’s why image recognition is such a good use case for machine learning.
Rather than programing a machine to look for certain dog-like characteristics, the engineers will instead create a program that will iterate based on inputs and output pairs. Humans will provide the algorithm with a large number of pictures as input, each with a corresponding output stating whether or not that image includes a dog.
With enough data and training, this iterative model can create startling accuracy. Before long the program will no longer require human training or examples. It will be able to identity dog pictures based on its training. Engineers will then test the program’s accuracy, make any needed adjustments and repeat the process until the algorithm is accurate.
Now that you understand how supervised learning works you can probably guess the difference in unsupervised learning.
In unsupervised learning the system iterates without the labeled, structured data that teams use to train supervised algorithms. Another way of putting this is that the algorithm has to train only on inputs without knowing the corresponding outputs.
Unsupervised learning algorithms attempt to model the underlying structure of data to understand it and predict outputs.
Machine learning vs. deep learning
You might have heard the term deep learning before, but what is the difference between machine learning vs. deep learning?
In the same way that machine learning is a type of AI, deep learning is a type of machine learning. The difference between deep learning and other types of machine learning algorithms has to do with something called a neural network.
A neural network is a system that loosely attempts to emulate the way a human brain solves problems. It does this using layers of connected units to learn relationships based on data. When there is more than one hidden layer in a neural network, the approach is called deep learning.Deep learning can be supervised, unsupervised or even semi-supervised.
People already use deep learning to solve some of the world’s toughest problems, including training self-driving cars and diagnosing cancer.
Now that we have a better understanding of some of these terms, let’s look at a visual of how it all comes together.
Why business leaders should care about machine learning
This technology has already drastically altered the business landscape. Let’s take ecommerce as an example. Picture a large online retailer’s homepage. The moment you open your account you receive highly accurate recommendations. This can benefit both the seller and the buyer. Could a traditional algorithm make such accurate and personalized recommendations for millions of customers?
Examples like this are endless. A car’s computer system can analyze millions of pictures to recognize hazards on the road. A weather system can ingest sensor information to predict the weather with astounding precision.
Other use cases include cyber security, health care, process automation and financial analysis. There’s a huge chance this technology has already impacted your operations. The question now becomes how your team will use it to succeed.
Hopefully this guide was helpful. If you are interested in learning more, download your free copy of Machine Learning for Dummies.