In the , we saw the first two types of machine learning. In this post, we will discuss the other two types of machine learning. These are — Semi-supervised machine learning and Reinforcement learning. previous post Semi-supervised Learning This type of learning is new and is an important part of the machine learning field. We use this type of learning when we have some labeled data and many unlabelled data in training. This type of learning is in between supervised and unsupervised learning. The part where we have the labeled data falls in the supervised domain, and the amount of unlabelled data falls in the unsupervised domain. Getting the labeled data is very expensive and time-consuming while acquiring the unlabeled information is easy and inexpensive. The use of unlabeled data with the labeled data produced a considerable increase in learning accuracy. Thus, situations may arise where we have a small amount of labeled data and a large amount of unlabeled data. Then in such cases, we can neither go to a supervised nor unsupervised learning strategy. So, in such cases, semi-supervised learning comes to our rescue. We can significantly increase our accuracy and save time and effort in labeling every data point. The analogy. In layman’s language or in words that are easy to understand, semi-supervised learning is like supervising a student for a short amount of time and then letting him go and wander the field independently. It solves classification problems. That means that you will need some supervised parts. Then at the same time, you have to train the model on large datasets of unlabelled data, for which you need the unsupervised part of machine learning. The central concept is to cluster different data points in the same clusters and then use supervised learning to name data points or instances in those groups. For example, suppose we have a data set of 1000 images of different vehicles. And we have four categories, and 100 data points or instances are labeled with those category names. We are then taking the concept of semi-supervised learning. First, we need to make clusters of images containing identical vehicles. Once the group of clusters has been formed using an unsupervised learning approach, then the task of supervised learning comes to play. We then assign the names to various collections and thus multiple instances in those clusters. In this way, we use semi-supervised learning to train the model using both learning strategies, i.e., supervised and unsupervised learning. Reinforcement Learning In this machine learning strategy, we are concerned about how an agent maximizes the rewards. There are some essential concepts in this learning, which need to be understood. The First — Agent. An agent is the action doer-the machine, which does the action in the environment. The whole aim of the agent is to make the rewards maximum by doing the activity right. If the agent does the action wrong, then the agent gets punished means, the reward is lessened from the total reward repository of the agent. The Second — Action. These are the actions that the agent does on the environment. Some examples are a robot walking, a robot playing tennis, etc. The Third — Environment. The environment is the playing field for the agent. It is where he does the actions. The environments can be different. For game-playing AI, it is the game. For the mopping robot, it the house, and for the mars rover, it the mars. The Fourth — Reward. The reward is an achievement, that increases the score of the agent. The learning process is continuously circulating the rewards. The agent tries to maximize the rewards, and it does the word so that its regards get up. An example of a reward can be, an increment in the score while playing the game. The Fifth — State. The state means the updated environment. When an agent does some action on the environment, the environment gets updated. That updated environment is returned to the agent and is known as the state. This usage of this type of learning is when we don’t have any data to start. In Reinforcement learning, the agent begins executing the tasks, and if he does the tasks correctly, the agent gets rewarded, and if he makes a mistake, he gets punished. Like in the above example, Increment in the score is a reward, and a decrement in the score is punishment. For an analogy, it like leaving a man in the car and telling him to figure out how to drive. He will learn the driving by himself by making mistakes and rectifying them. Here, the reward can be money. If the driver does not make any mistake, he will receive 1$, and if he makes a mistake, he will give 50 cents. Conclusions These are the four types of machine learning strategies. All the machine learning algorithms fall into one of the above Learnings. They offer a vastly considerable number of algorithms that can be implemented. We will discuss them in future posts. If you liked the content, please share 🙂 , and if you think the content needs to be improved, please comment below with suggestions. Also Published Here

Learn and Share Ai, Machine learning and Deep Learning.

2022 - HackerNoon Contributor of the Year - Deep Learning

Come and let us finish the AI threat together.

Read My Stories

Nominated for 2022 - HackerNoon Contributor of the Year - Deep Learning

Too Long; Didn't Read

In your car, at home, or at work — Bosch technology shapes many areas of life.

The Four Types of Machine Learning | Part 2

Too Long; Didn't Read

M Shehzen Sidiq

About Author

TOPICS

THIS ARTICLE WAS FEATURED IN...

The Four Types of Machine Learning | Part 2

Too Long; Didn't Read

M Shehzen Sidiq

About Author

TOPICS

THIS ARTICLE WAS FEATURED IN...

RELATED STORIES