Hi, I'm Jaekwang Kim from Sungkyunkwan University. From now on, we'll talk about the basic concept of machine learning. This is my first time, so I want to talk about just the two things. One is what is machine learning? And the other is the three types of machine learning, so share with that. And this is the whole content of today, so I want to first talk about what is machine learning? And then I will share about three types of machine learning such as supervised learning, unsupervised learning and reinforcement learning. So what is machine learning? Let's just talk about the artificial intelligence. It's the science and engineering of making intelligence machines, and machine learning is a branch of artificial intelligence, so it's a sort of subset of artificial intelligence. And deep learning it was a famous concept, and it's also the subset of machine learning. Actually it is based on the neural network, which is one of the machine learning algorithm. So I want to talk about what is emotional learning. I can tell you the machine learning is to improve the performance or programs based on given data, previous regional or experiences. The main difference with the traditional programming is this in traditional programming data. And rules are used as input and it is used the input to the computer program and it makes the results as an output in motion learning. On the other hand, the result and data is huge, their input to the computer program and it makes the rules as the output. So why we should learn or concerned about motion learning? First, because machine learning is a tool to reduce programming time. And second, machine learning personalize your product to provide what's better for specific group of users. So it provides a solution when on algorithm to solve a problem is not available or difficult left but not the list. It allows people to think at a higher level rather than focus on simple task. Here are the three types of machine learning, it's supervised learning, unsupervised learning, and reinforcement learning. Supervised learning uses the label data and has a direct feedback, and the purpose of the supervised learning is predict outcome and futures. Unsupervised learning had no rabels, no direct feedback. The purpose overrules provide learning is find hidden structures, how about the reinforcement learning? It's a decision process and it has a reward system and the aim of it is long fridge of actions. And here are the big pictures of the machine learning and three types of machine learning and all of which problems. First of all, in the supervised learning there is two representative problems, classification and regression. And there are many specify then and the problems like here and in also provide learning. There are two main problems such as dimensional reduction and clustering and meaningful compression and structure discovery belongs to dimensional key reduction. And we can also use the clustering for recommended system, targeted marketing, and so on. How about the reinforcement learning? It is constant with the game AI and skills acquisition and something like that. And first of all, let's deep dive into the supervised learning. And I can say supervised learning is how to train a capital program with labor data that is given explicit correct answers. And it's about how to proceed with learning in the form with data and labor and in superbad learning. Let's take our amnesty data set as an example. It's 28 times 28 images which is like this the training set conscious of the following. And there are 10 labors from 0 to 9 for example we have a 20 by 20 images representing Gerald. And it also had the label Gerald, and we also have about 20 times 28 images representing seven, and it had the label seven, and we also get some the other data. If we we input the image 4 to the trend model with amnesty dataset, we can finally get the predictor label as of 4. In an inspired by the learning, I can say it's no label for data. That is a methodology that trend or comp it'll without the explicit correct answer. So it's a method of running form like with only data we can simply say with these figures in case of supervised learning. The proposal of classification is making a decision boundary like this, and how about the unsupervised learning? The representative method is clustering like this, so we want grouping these two groups. The less the thing is reinforced much learning reinforcement learning is not supervised learning. It's not all supervised learning in reinforcement learning agent to learn to maximize reorder. And it's an algorithm that includes the process of collecting data in kind of dynamic environment. So if the agent make some action at a time t, and the environments returned reward and state at a time step t plus 1. So agent makes you want to make the maximize the reward with this environment. So I just say about three types of machine learning like this which of the best time method algorithm of emotionally. I don't want to say the best algorithm for the every problems, but I believe that you can choose the best algorithm for your case of problems, so that you may make the right decision. Thank you.