AI models
1. AI models
SPEAKER: Before you dive in to more practical topics and build your own machine learning model, let's prepare you with foundational knowledge and explore the AI model categories. First, let's pause to clarify two terms, artificial intelligence and machine learning. You may note that people often use the terms interchangeably, but they do have some differences. Artificial intelligence, or AI, is an umbrella term that includes anything related to computers mimicking human intelligence. Some examples of AI applications include robots and self-driving cars. Machine learning is a subset of artificial intelligence that allows computers to learn without being explicitly programmed. This is in contrast to traditional programming, where the computer is told explicitly what to do. Machine learning mainly includes supervised and unsupervised learning. You might also hear the terms deep learning or deep neural networks. This is a subset of machine learning that adds layers in between input data and output results to make a machine learn at much depth. You'll learn more about neural networks and deep learning later in the course. Finally, generative AI, or GenAI, creates content and performs tasks based on requests. GenAI uses foundation models like large language models, a type of deep learning model, to predict, interpret, and interact with language. You'll delve deeper into GenAI models in the next module. So what's the difference between supervised and unsupervised learning? Imagine two types of problems. In problem one, you are asked to classify dogs and cats from a very large set of pictures. You already know the difference between dogs and cats, so you label each picture and pass the labeled pictures to a machine. By learning from the data, in this case, pictures with the answers or labels, supervised learning is being enacted, allowing the machine to tell if a new picture represents a dog or cat in the future. In problem two, you are asked to classify breeds of dogs. Unfortunately, this time, you don't know many of them and are not able to label the pictures. So you send these unlabeled pictures to a machine. In this case, the machine learns from the data without the answers and finds underlying patterns to group the animals. This is an example of unsupervised learning. Put simply, supervised learning deals with labeled data, is task-driven, and identifies a goal. Unsupervised learning, however, deals with unlabeled data, is data-driven, and identifies a pattern. An easy way to distinguish between the two is that supervised learning provides each data point with a label or an answer, while unsupervised learning does not. There are two major types of supervised learning. The first is classification, which predicts a categorical variable, such as determining whether a picture shows a cat or a dog. In ML, you use models like a logistic regression model to solve classification problems. The second type of supervised learning is regression, which predicts a numeric variable like forecasting sales for a product based on its past sales. You use ML models like a linear regression model to solve regression problems. There are three major types of unsupervised learning. The first is clustering, which groups together data points with similar characteristics and assigns them to clusters, like using customer demographics to determine customer segmentation. You use ML models like k-means clustering to solve clustering problems. The second type is association, which identifies underlying relationships like a correlation between two products to place them closer together in a grocery store for a promotion. You use association rule techniques and algorithms like Apriori to solve association problems. And the third type of unsupervised learning is dimensionality reduction, which reduces the number of dimensions or features in a data set to improve the efficiency of a model, for example, combining customer characteristics like age, driving violation history, or car type, to create a simplified rule for calculating an insurance quote. You use ML techniques like principal component analysis to solve these problems. All right. Time to test your learning. You are asked to predict customer spending based on purchase history. Is this supervised or unsupervised learning? Yes, that's supervised learning because you have the labeled data, the amount the customers have spent, and you want to predict their future purchases. Is this a classification or regression problem? Yes, it's a regression problem because it predicts a continuous number, future spending. Which ML model should you use? A logistic regression or a linear regression? Yes, a linear regression. A logistic regression model is for classification problems, while a linear regression model is for regression problems. Let's look at another scenario. Imagine you were using the same data set. However, this time you were asked to identify customer segmentation. You don't want to base your judgment on stereotypes such as age or gender. So you use a computer for help. Is this supervised or unsupervised learning? Yes, it's unsupervised learning because you don't have each customer labeled as belonging to a certain segment. Instead, you want the computer to discover the underlying pattern. Is it a clustering, association or dimensionality reduction problem? Yes, identifying customer segmentation is a clustering problem. Which ML model should you use? A logistic regression? A linear regression? Or a k-means clustering analysis? Right, it's a clustering analysis scenario. You will find these models within BigQuery ML, AutoML and Custom Training later on in this course.2. Let's practice!
Create Your Free Account
or
By continuing, you accept our Terms of Use, our Privacy Policy and that your data is stored in the USA.