Understanding Machine Learning

You've probably heard the term "Machine Learning" pop up lately. Machine learning is becoming more common in our daily lives, from movie recommendations on Netflix to virtual assistants like Siri. But what exactly is it? Let's break it down.

What is Machine Learning?

Machine learning describes software that can learn how to do things, like a software program that can identify flowers or dogs and tell you what kind it is. To learn about machine learning, it will help to understand how a standard software program would work and compare it to a machine learning program. First, let's look at a non-machine learning program created to identify dogs from a picture. To start, we would take a few thousand photos of dogs and create a database of "dogs." Our program then is instructed to match any new photo of a dog with one in our database, and if it matches, to let us know that the new image is a dog. 

When we pointed a camera at a picture of a dog, the program would try to match the image with one already in its database. If there isn't an exact match, the picture we showed isn't recognized as a dog. While this program is helpful, it can only recognize photos that match those already in its database. What would be really useful is a program that could identify a dog in a picture even if there wasn't an exact match in the photo. That's where machine learning comes in.

How could we teach a machine?

Instead of programming a computer with explicit instructions to recognize a dog by comparing pictures in a database or to look for specific items in an image like a tail, four legs, and a nose, we use a machine-learning program to "learn" what a dog looks like. 

We start and show it thousands of pictures of dogs. By examining the pictures, the program is trained to learn what features are commonly associated with dogs. It forms its own understanding of what a dog looks like as it examines images of dogs. Instead of relying on a specific picture match or instructions to look for a tail, ears, and four legs, the machine learning program has formed a concept of what a dog looks like and can recognize one that it hasn't seen before. 

Even if the machine learning program has never seen a picture of a dog before, it uses experience to analyze it, then makes decisions to see if it has the characteristics that make it a dog or something else. 

Machine learning involves teaching a software program, the "machine," to improve at whatever task we want, like recognizing a dog in a picture. To become better at the job, Machine Learning has to go through some steps and have some capabilities that allow it to learn.

Let's look at the parts of how Machine Learning learns something.

How Does Machine Learning Learn?

1. Training Data: This is the machine's textbook or reference information. If we return to the dog example, the training data would be all the dog pictures we initially showed the program. This helps it form a basic understanding of what a dog looks like because it analyzes the images to learn about them. The more images of dogs we give it to analyze (the data), the better the machine learning becomes. Machine learning requires providing the program with as much data as possible since more data results in more experience. 

It's like a writer who reads as much as possible to expand their vocabulary. The writer doesn't need to memorize what they read, but can learn new words, their meaning, and then form new sentences. 

2. Model: Think of this as the machine's brain. Based on the training data, the computer creates a model, like a prototype, to understand the data that it's been given. For instance, it might recognize that most dogs have noses of a particular shape or fur of various kinds. By examining the data, the machine learning program discovers the similarities and differences in its data, like the dog pictures, and can learn to recognize photos of dogs it hasn't seen before. 

3. Testing: Like a pop quiz, we show the program new pictures (ones it hasn't seen before) to see if it can correctly identify a dog. If it makes mistakes, we can tell the program that it made a mistake, or got the answer correct, and train it further. People might participate in the initial testing to help the machine learn better. The human testers can help the machine understand its correctness and enter a yes or no to assist the program. As the testers enter feedback, the machine learns what to do the next time it comes across a new situation, like a picture that confused it earlier, and improves.

4. Application: Once we are satisfied with its learning, we can use the machine's knowledge in real-world scenarios. Now, when it sees a new image we show it, it can tell if there's a dog in it or not! 

Even after this, the machine (software program) can be improved with feedback. If it makes a mistake or provides a correct answer, people who use it can let it know, and it continues to learn.

Why Is Machine Learning Important?

The beauty of machine learning is its adaptability. Since it learns from data, it can get better and more accurate over time as it is exposed to more data. Traditional computer programs don't have this ability. Once they're written, they follow the same set of instructions repeatedly. 

Think of the program we described earlier that would identify a dog based on trying to match the picture you show it to an image it had in its database. If there wasn't a match, it would tell you that the picture was not a dog. Machine learning bases its answer on what it has learned about what a dog looks like, not on any one specific image. 

The reference material or data that machine learning uses to solve problems can change. Imagine a new breed of dog that the program could learn to recognize, or to recognize a dog from the front, back, or side, or when it's jumping or swimming. But machine learning can do more. 

Machine learning helps with something we already use, speech recognition. Our accents, the words we use, or even background noise can vary wildly. Machine learning systems can adapt and learn from these variations, making them invaluable in today's rapidly changing digital world.

With a software program that adapts, it's possible to use it for solutions that involve unpredictable situations and changing information without creating a whole new update for the software. In effect, the software updates itself or "learns."

Here are some ways Machine Learning is already used.

Applications in Everyday Life

• • Personalized Content: Ever wonder how YouTube or Spotify seems to know your tastes so well? Machine learning analyzes what you're watching or listening to on a streaming service and recommends content that matches your preferences.
• • Email Filtering: Those pesky spam emails? Machine learning helps filter them out, ensuring your inbox remains clutter-free.
• • Virtual Assistants: Siri, Alexa, and Google Assistant get better at understanding your requests over time, thanks to machine learning.
• • Medical Diagnoses: Machine learning can assist doctors by analyzing medical images and detecting signs of diseases often too subtle for the human eye. 

Wrap-Up

Machine learning might sound like a fancy tech term, but at its core, it's about teaching computers to learn from experience. Just as we improve at a task with practice, machines can do the same. The more data they're exposed to, the better they become at their jobs.

As technology evolves, machine learning will play an even more significant role in our lives, making our gadgets smarter, our lives easier, and the world more connected. Human beings create tools to make jobs easier, like hammers. Machine Learning is just another tool that humans have invented to solve problems. 

So, the next time you get a movie recommendation or your email filters out spam, take a moment to appreciate that machine learning has a massive impact on your life!

Check the calendar for new Tech Topics, internet marketing, social media marketing, eCommerce, technology, and other classes.