I’m Kesha Williams, an AWS Machine Learning Hero and Alexa Champion, and this is Kesha’s Korner, where we learn about artificial intelligence and machine learning on AWS. Learn with me as we transform your engineering skills and future-proof your career!
What is Machine Learning? ML is a hot topic!
So here’s how it works.
A learning algorithm studies data, looking for trends and patterns. The trends and patterns it finds are then stored in a mathematical model. You’ll hear data scientists talking about the “model”. That’s what they’re referring to, the trends and patterns found in data. And once you have this model, it can be queried or consulted by giving it new data that it hasn’t seen before and then the model provides a prediction.
Let’s take the case of using machine learning in healthcare. There are some machine learning models that can accurately predict cancer long before a diagnosis. That’s mind-blowing, but how does it work?
First, the model undergoes a training (or teaching) process where the learning algorithm studies images of cancerous tumors and non-cancerous tumors, and it learns to identify what a cancerous tumor looks like. So when it is presented with a new tumor, it can predict (or “infer”) whether or not this new tumor is cancerous.
So, how do machines learn? There are several machine learning techniques.
The first technique is called supervised learning. This is where the algorithm learns from labeled data. Going back to our healthcare example, an algorithm is fed images of tumors that have been labeled by a human. So Image A is labeled “this is cancer”, Image B is labeled “not cancer”. Later on, we’d test our trained model on new pictures that have not been labeled.
The next technique is called unsupervised learning. This is where the algorithm learns without labeled data, so the algorithm tries to find patterns that govern the data. It tries to organize or cluster or group data based on common features it identifies.
The third technique is called reinforcement learning. This where the algorithm learns through consequences of an action; the algorithm receives feedback as it’s learning. We tell the machine: “this was a good decision, here is your reward”, or “this was a bad decision, here is your negative consequence.” You see this a lot in robotics.
Want to learn more from Kesha? Check out her Certified Alexa Skill Builder Specialty Exam course on A Cloud Guru!
What type of questions can a machine answer?
First, machines can answer Yes or No questions. This is known as binary classification. Examples of these questions might be: Is this tumor cancerous or not? Is this email spam or not? Is this transaction fraudulent?
Machines can classify data with more than two options as well. That’s called multi-class classification. For example, is this image a picture of a cat, dog, or a bird?
Next, machines can provide a numeric answer, known as regression. (What will the temperature be in Atlanta tomorrow? How many units of this book will sell? )
Machines can also group data, which is called clustering. If you have large quantities of customer data, a clustering algorithm might reveal distinct groupings for your customer base, and thereby help you understand who your customers are.
Machines can even generate something new based on existing data! We call that generative AI, and it’s great at coming up with new samples or designs from a pattern. For example, given an airplane part, generative AI could help us refine designs to make the aircraft lighter, stronger, or more efficient.
Machine learning is the next wave of transformative technology that will change life in unimaginable ways. Today we looked at what machine learning is, how machines learn, and the types of questions a machine can answer. I hope you’ll join me in our next edition of Kesha’s Korner, where we will look at the ML lifecycle and how it differs from the more traditional software development lifecycle (SDLC).
Ready to become a machine learning engineer? A Cloud Guru’s AWS Machine Learning Specialty Certification course has your back.