Suppose you want an easy understanding of machine learning and its algorithms and also various other things in machine learning. But you could not get those explanations from anywhere. You are not alone; I myself was one of the new learners and even am learning today. And I am also trying to improve my knowledge in this field. If you know machine learning only by definition, you are ready for these learning strategies that we use to train machines. So let us talk about the types of machine learning that we use daily (like Machine Learning students, Machine Learning Engineers, etc.) But if you have no idea or have forgotten, then have a look at previous posts.
This will be a two-post article. In the first, we will go through the first two types, and in the next, we will learn the next two. Now, as we are all set, let us dive into the good meet.
There are three types of machine learning. Initially, there were three, but later type added one more type to the ranks of machine learning types. Thus in total, we have 4 types of machine learning strategies. Let's see what those are and then define them one by one. We will go into the brief depth of these strategies to understand their upper creamy layer of them. When we are ready to learn them in-depth, I will be posting each algorithm that comes in these strategies. When we know the algorithms in detail, then we will be ready to implement them. Let me know if you want the practical implementation also using Python.
Let us see what different types of strategies that we have been talking about are.
From the above diagram, we can see that there are FOUR types of machine learning. These are as under:
Supervised machine learning is the learning strategy that maps an input to the output based on example input-output pairs
Now, let us try to understand this most thoroughly. You remember the situation where you are given a question paper and an answering script for practice to prepare for the examination. You had to practice and come the next day prepared for the examination. The same happens here; in our case, the questions and answers were input and output, respectively. In the same manner, supervised learning involves giving the algorithm both questions and answers for learning. Now another concept comes into play. That concept is known as “Teacher.” The person who corrected us when we were wrong. In the case of machine learning, we may say that there is a supervisor. This supervisor is supervising the whole process of learning. When the algorithm makes a mistake, this supervisor corrects the mistake or at least moves closer to the correction.
In a nutshell, we may say that supervised learning is that learning, where the learning algorithm is given both the answers(Output feature) and questions(input features). The algorithm learns the mapping, and then once the model has learned this mapping, it is tested with the data it has never seen. This testing gives us an idea, How well the algorithm has learned and can deal with the new data? This concept is known as Generalization (which we will talk about in coming articles). This testing is similar to us appearing in the exam and being tested with the questions we have not seen while preparing.
Supervised learning can be classified into two more subtypes:
Supervised machine learning subclassification
I will try to touch on these concepts, and we will talk about them in detail in the future.
Regression, in simple terms, is when we have to predict a continuous value, like predicting salary based on experience, age, etc. In regression, the basic concept is to fit the line in the data such that the sum of all the differences between the points and the line is minimum. This line-fitting concept is the same as what we have studied in high schools.
In these problems, the algorithm tries to predict the class of the input. This is the most common application of machine learning. It is like saying whether it is a cat or a dog when shown as a picture of one of them. Classifying animals into their habitats, classifying the emails as spam and not spam, etc. There are many more examples in this.
Unsupervised is different from supervised machine learning because we don’t have an outcome or output variable. The data we have been given is not labelled, which means we have not been told which one is a cat or a dog. In machine learning, we feed the unlabelled data to the machine, and it tries to find the similarities between the data points and comes with the clusters(in simplest form). The points in these clusters are similar, based on the characteristic which our algorithm has found.
If we talk about the analogy, it is like giving a child bunch of toys and telling him to separate the similar toys into their respective groups.
Unsupervised learning has further 2 sub-types:
Clustering is what we talked about. Making the groups of similar things or, in machine learning terms making groups of similar data. In the below example, our algorithm learns the differences between circles, squares, and triangles. Thus making clusters of them.
Associations mean finding the relations between various data points and associating them with each other. Malls, Walmart, etc., use this technique. They use this technique to recommend the things that are bought together to various customers. Let us take an example of parents buying Diapers for the baby, and it has been seen that those who buy diapers also buy baby food. This relation is learned by the association algorithms and recommends the products to parents based on these associations.
This completes part 1 of the article. Be connected for the second part. If you like my work, share this with your machine learning friends :).
This article was first published here.