From the course: Artificial Intelligence Foundations: Neural Networks
Machine learning and neural networks
From the course: Artificial Intelligence Foundations: Neural Networks
Machine learning and neural networks
- [Narrator] We begin by answering the question, "What are neural networks?" After completing this chapter, you will be able to describe the differences between machine learning and neural networks, understand how biological networks lay the foundation for artificial neural networks, and describe the single-layer perceptron. In order to answer the question, "What are neural networks?" we need to define machine learning. We begin with an example. One question that machine learning can help with is predicting some type of outcome. For example, you are studying for a certification exam. You ask yourself, "If I study really, really hard and if I put some serious study hours, can I improve my score on my certification exam?" This is a prediction problem that can be solved using machine learning. You want to know if your future exam score can be predicted based upon the number of hours you study. This image shows a graph of hours of study on the x-axis versus test scores on the y-axis. Let's assume this data represents prior test takers who were surveyed about the number of hours they studied. Test takers who studied between five and 10 hours received test scores between 50 and 60 and failed their certification exam. Test takers who studied between 30 and 35 hours received test scores close to 90 and passed the certification exam. So yes, it appears you can improve your test scores based on this pattern, meaning there is a positive relationship between hours of study and test scores. Machine learning uses data to detect various patterns in any data set by analyzing and outputting a prediction. The machine learns from past data and improves it automatically. For example, a machine learning mathematical formula, or model, takes the input data, the table shown here, and learns the relationship between the number of study hours and test scores. In machine learning speak, we try to map the relationship between input data and output data in such a way that it can generalized or be applied well on unseen data. For example, Exam Day 1 is our original input data. Exam Day 2 and 3 are the unseen data. Our machine learning model will be able to apply the learned relationships to this unseen data. Based on our machine learning these relationships, your future test score could be predicted. (chuckles) In fact, the more data you feed it, the more it's learning improves. Training of the model takes place by comparing the actual output with the predicted output. A cost function is used to reduce the total error between the actual and predicted values. We will learn more about cost functions in later videos. At a higher level, machine learning is the branch of artificial intelligence, or AI, which is devoted to making machines think and act like humans. In our certification exam use case, our exam machine would have the ability to automatically learn and improve from experience if we actually used an ML framework to test the model. The primary aim is to allow the computer to learn automatically without human involvement or assistance and then adjust actions accordingly. A subset of machine learning is called deep learning, which is based on artificial neural networks, which are based on biological neural networks. Let's take a look at that next.