The term machine learning is often used in a variety of ways, some of which aren't entirely accurate. In this video, learn exactly what machine learning is.
- [Instructor] This lesson is going to focus on the seemingly elementary question, what is machine learning? We're starting here just to make sure we're going into the rest of the chapters with the same understanding. If you ask 10 data scientists to define machine learning, you might get 10 different answers. But they will revolve around some key themes. So let's explore a few of those definitions. The first definition comes from Arthur Samuel. Samuel is recognized as one of the first real machine learning pioneers. And he was actually the first to coin the term, machine learning. He defined machine learning as the field of study that gives computers the ability to learn without being explicitly programmed. This is a pretty good definition, but it's a bit vague and seemingly magical. So let's try to get a little more concrete. This definition from the University of Washington hits on one key concept that is missing from the previous definition. Machine learning algorithms can figure out how to perform important tasks by generalizing from examples. Again, generalizing from examples. That's a very important point, but I'll revisit in just a minute. So borrowing that portion from this definition, here's my aggregation of several definitions that I think makes things just a little bit more concrete. Machine learning is fitting a function to examples and using that function to generalize and make predictions about new examples. This hits on the fact that algorithm, or machine learning model, is based on the data that you feed it that's learning from examples and that the entire goal is to use that learned model to make predictions about new examples. As I mentioned before, this is a really key concept. In other words, machine learning models learn from trans and past data to make predictions about future data. If you think about it, we all do this on a day to day basis. We learn from our past experiences to adjust our behavior or our views in the future. With that definition in mind, an even simpler definition of machine learning is simply pattern matching. Again, a model learns from a pattern and data that was fed to it, fits a function to that pattern, and then uses that function to pick up on those patterns in future data to make predictions about it. Let's look at a really simple example of machine learning. So this is a very simple, single variable linear regression. So this plot is showing the number of umbrellas sold based on the amount of rainfall. So the model seeks to predict how many umbrellas will be sold based on the amount of rainfall. The model in this image is just represented by this red, best fit line. You might remember that the equation for a line is just y equals mx plus b. Where y is the thing that you're trying to predict, which is umbrellas sold in this case, x is a thing that you're trying to use to predict it, that's rainfall in this case, m is the slope of your line, and b is the y-intercept. So this best fit line on this plot has an actual equation. The equation of that line is your model, or it's the function that is fit to this data. Remember back to our definition. Fit a function to examples and use that function to generalize and make predictions about new examples. So we covered that first part, we have a function or a model that is fit to data, that's this line, now what does this model allow us to do or say? Well, if it happens to rain 110 millimeters, even though I don't have any examples of days where it rained exactly 110 millimeters, I can say that based on my model, we could expect about 30 umbrellas to be sold. So this is a very simple example of the machine learning model.
- What is machine learning (ML)?
- ML vs. deep learning vs. AI
- Handling common challenges in ML
- Plotting continuous features
- Continuous and categorical data cleaning
- Measuring success
- Overfitting and underfitting
- Tuning hyperparameters
- Evaluating a model