Types of machine learning

And the first branch of artificial intelligence is machine learning. So machine learning is the science of programming computers. So they can learn from data means that you provide the information, the data to computer, and then they can learn from the data. And to give a definition of machine learning, I try to provide two different definitions here. For the first definitions, from Arthur Samuel, from 1959, he defied machine learning is the field of study that gives computers the ability to learn without being explicitly programmed.

And the second one is more algorithm from Tom Michelle, 1997, a computer program is set to learn from experience E with respect to some dusty and also some performance measured P If its performance on T as measured by P, so improves with experience E. So there are the definitions of machine learning. And why we need to use machine learning ? So you can see in this figure. This is the difference between a traditional programming and machine learning programming. In traditional programming, if you want to study the problems. You need to write some rules, and after that, you evaluate and launch the applications. During the implementations, You can have some analysis on the errors.

If you have errors, try to analyze and then implement them again. However, what is the difference with machine learning here? So in machine learning, you replace the white “write rules” mean that you don’t need to apply any rules. You need to write any rules and the computer can understand. And how they can understand; How they can study the problems. So we need to provide the data. So they can learn the data. And from the data, they can use this to train machine learning items. And after that, the other process is similar. They try to evaluate according to the model that they have trained. And finally, they launch the applications, they also include the analyze error step during the implementations.

That is the difference between machine learning and traditional programming. And it means that in machine learning, if you give computers a lot of data. They can learn and then train the algorithms, very well. Because currently, there are a lot of data nowadays, in different fields. That’s why machine learning currently become very popular. And in bioinformatics also is popular also. And what are many types of machine learning? I try to show you, as I already explained. In the first, like about some types of machine learning in AI. So here more detail on machine learning, there are some kinds of machine learning algorithms, and machine learning techniques.

So the first, and we can divide machine learning in two, in three different main process main techniques. The first one is supervised learning. Sescond one is unsupervised learning. And the last one is reinforcement learning. For supervised learning, there are two type of two popular tasks. We call classification. And the second one is regression. Classification, mostly for some of the classifications problems, like face recognition are also, like the diagnosis of some disease are in… However in regression, the label, the output, the label is the number. So you try to predict the number like you try to predict the weather, or you try to predict the price of a house sometimes. And the second one is the unsupervised learning.

For unsupervised learning, this is the types of machine learning to help you to reduce the dimensions of… in bioinformatics, when you try to perform clustering techniques, you need to use the unsupervised learning. And the last one, of… last type of machine learning is reinforcement learning. And for this one, mostly used in developing some robots and because they can learn from the environment. So let’s go to more detail in supervised learning. In supervised learning, means that the training data, you feed to the algorithms includes the desired solutions. We call them label, so means that you need to provide labels to the computer and the computer can learn according to this label. And there are two kinds of supervised learning.

As I mentioned, first one is classification and the second one is the regression. And for example, in this figure, if you want to create a robot that can help you to identify… to classify the fruit like the apple, here. If you try to input a raw data, “this is the apple.” And then the computer can use machine learning to learn the information to learn all of the characteristics or the features of the apple.

After that, they can have a model, and the model will be used to predict the outcome, which means the future, if in the future, you insert a fruit with us, with some characteristics into our machine learning model and robot, our build… our creative robot can help you to classify with this one belong to apple or not. So this is an example of machine learning for classification. And difference with the classification machine learning, we have a regression machine learning. And what is the regression machine learning? Here, because it is also supervised learning, so you still have labels for that. However, the label, here is not a binary label.

The binary labels means that you classify there are apple or not, or they are which types of food. However, here, they try to predict like the age of a person which means the label is continuous number. And then, another example is, for example, you try to predict the stock price of a company, and also the predict the medical costs. So for all of these problems, we call regression which means the label is continuous.