Welcome to the fourth and final course of this specialization. You're just one course away from finishing this whole specialization and having learned a lot about TensorFlow. In this course, you'll learn about sequence models. What are sequence models? So, what we're going to be focusing on is one part of sequence models, which is really time series. Sequence models, where it's a case of, if you can imagine a series of data that changes over time. It might be the closing prices for stock on the stock exchange, or it could be weather. It could be how sunny it is in California on a given day, or how rainy it is in Seattle on a given day. That type of thing. So, if you just imagine how an item of data changes over time, and how it's measured over time. So, basically, almost the same thing, like a spreadsheet. If I have a spreadsheet where I have one day per row, and two columns, say, one for the California weather, one for the Seattle weather, to document how better my life is than yours living in Seattle, living in California. How wet are mine is. Yeah. Then we would have a neural network help us model that. Exactly. Exactly. So, we're going to start by creating a synthetic sequence of data, so that we can start looking at what the common attributes that you see in data series are. So, for example, weather data can be seasonal, right? It's sunnier in June than it is in January, or it's wetter in November than it is in October. Something along those lines. So, you have that seasonality of data. You can also, in some cases, have trends of data. Weather probably doesn't really trend, although we could argue that it's trending upwards. A little bit of climate change. Yeah, with climate change. But stock data may trend upwards over time, or downwards over some other times. And then, of course, the random factor that makes it hard to predict is noise. So, you can have seasonal data, you can have trends in your data, but then you can have noise on that data as well, so that the average temperature of a Tuesday in June in California might be 85 degrees, but it might be 85.5 degrees, it might be 84.5 degrees, so you get that kind of noise in the data. So, we want to start looking at various methods that can be used statistically and with machine learning to help us predict data given seasonality, trend, and noise. Cool, great. And then, in this course, at the end of this course, one of the most cool applications is to use these ideas to model sunspot. Sunspot activity, yeah. So, sunspot activity is really interesting because the sun has like an 11-year cycle, although some astronomers tell me it's a 22-year cycle, there's actually two 11-year cycles like nestled beside each other. And will we resolve this in this course? Ah, that remains to be seen. You're going to have to study all the way through, and then we'll see. But the idea then is that you do get that nice seasonality, and we have data measuring back about 250 years' worth of sunspot activity, so that's on a monthly basis counting the number of sunspots that have been spotted by astronomers. So we do definitely see that 11-year cycle, or maybe the 22-year cycle, and there's a lot of noise in there, the seasonality, and that kind of stuff. So it's kind of fun to build something to predict sunspot activity. Yeah. And in fact, sunspot activity is very important to NASA and other space agencies because it affects satellite operations. So in this course, you start by learning about sequence models, time series data, first practicing these skills and building these models on artificial data, and then at the end of this course, you get to take all these ideas and apply them to the exciting problem of modeling sunspot activity. So let's get started. Please go on to the next video.