Products and technologies that use machine learning are all around us. We're using them every single day whether we realize it or not. And in this guide, I want to walk through some of the basic examples of machine learning agents that are out there to help give us a frame of reference for how machine learning can be used.
Behavioral targeted advertisements
In our first example we will discuss behavioral targeted advertising and how machine learning is applied in business-to-consumer marketing. In an effort to increase efficiency, companies have adopted data management platforms that create a personalized User Profile for consumers. Within each profile, a variety of data is collected by way of cookies or purchased from second and third-parties. The behavioral data is then applied to a series of algorithms that place similar profiles in specific target market groups. Based on the group’s demographics, companies can utilize predictive modeling to formulate ads that align with specific consumer preferences, while reducing messages they may find irrelevant.
Self driving cars
One of the coolest, yet more complicated applications of machine learning involves self driving cars. One of the main tasks for machine learning is the continuous rendering of the car’s surroundings and the prediction of possible changes to those surroundings. Simply put, the car needs to get you to your destination without crashing. In an effort to “see” where it’s going the car uses sensory inputs from cameras, radar, and lasers. Machine learning then generates its own digital map by predicting what and where objects are, as well as where they will be. This constant influx and processing of information is just a chunk of the decision making process every self driving car needs to make to get you somewhere safely.
Sports statistics
For those of you who may not like sports, please forgive me. But, I can’t think of a more perfect pairing for machine learning than sports statistics. Individual and team success has always been evaluated by data: win percentage, batting average, field goal percentage, and so on. With the development of machine learning these key metrics have been broken down to more specific variables to help optimize results. Machine learning doesn’t stop at optimization, it can also help with dynamic shifts in strategy based on the opponent's behavior. With all that being said, you can see why this is probably my most favorite application of machine learning.
Spam filters
Spam filters are one of the more common uses of machine learning, and have been implemented for over two decades. Compared to self driving cars, the method of spam filtering is fairly simple. But in my opinion, it’s equally as clever. How it works is that emails are run through a classification processes that acts as a filter. The naive Bayes algorithm looks at every email and determines if its spam, or not spam, based on structure and word usage. I’d also like you to keep this example in the back of you mind. Because, we will be going into greater detail when we get to the Naive Bayes section of classification.
Gaming Facial recognition
Facial recognition may become a major contributor in gaming development. Many of the machine learning algorithms used for image classification in self driving cars can be applied in a similar fashion to personalize your gaming experience. Using changes in facial landmarks, machine learning can recognize emotional variations. Over time your console can learn what combinations of sensory stimuli elicits the strongest emotional response. It will have the ability to promote or inhibit your entire emotional spectrum based on your desired user experience. And who knows? Maybe your console will learn enough about you to kick you out of a game, and lock your account before you get mad enough to throw your controller!
Product recommendations
Another great example of real world machine learning technology are our recommendation engines. Amazon was one of the earliest proponents to integrating machine learning into their e-commerce platform. They built in the ability to track purchases on a historical basis and then they had that purchase history dictate which items would be recommended to other customers and they were able to generate much higher sales because they were able to look to see what they think someone might want to purchase based on their own historical data.
