Unsupervised, Semi-supervised And Reinforcement Machine Learning || Beginner Guide To Machine Learning
In the last section, we have seen What is Machine Learning and what is Supervised Learning. Now let’s see about Unsupervised Learning, Semi-Supervised Learning, and Reinforcement Learning. All of them are largely used, as we see them from an industrial point of view.
What is Unsupervised Learning
Unlike supervised Learning, In Unsupervised learning the training data is unlabeled. Here the system tries to learn without a teacher.
For example, Most of us use the google news feature, What google does is cluster the news of a similar kind and provide us in the news feed. Here Google uses the clustering algorithm of unsupervised learning.
Another example, Let suppose you own a mobile application. You want to run a clustering algorithm to try to detect groups of similar visitors. You notice that 60% of your male visitors love comics books. Sci-fi lovers are just 15% who generally visit at weekend. Similarly, you can cluster them into a group.
Some most important unsupervised learning algorithms
Clustering
K means
DBSCAN
Hierarchical Cluster Analysis (HCA)
Anomaly detection and novelty detection
One-class SVM
Isolation Forest
Visualization and dimensionality reduction
Principal Component Analysis (PCA)
Kernel PCA
Local Linear Embedding (LLE)
t-Distributed Stochastic Neighbor Embedding (t-SNE)
Association rule learning
Apriori
Eclat
What is Semisupervised Learning
In supervised learning, we got labeled data. But in unsupervised learning, we got unlabeled data. But in some cases, we got some partially labeled data, as labeling of data is time-consuming and costly. Some algorithm deals with the data that’s partially labeled. This is called semisupervised learning.
For example, We all are using some photo hosting service, like Google photos, or maybe any other. Let suppose you upload a group photo of 10 peoples. Google recognizes 6 of them as they match with some other photos. But Google doesn’t know about the rest of the 4 people. It means that here 6 people are labeled and 4 people are don’t. You can easily label the rest of the four people so that you can search them in the future by their names.
Most of the semisupervised learning algorithms are combinations of supervised and unsupervised algorithms.
What is Reinforcement Learning
In this type of learning two types of words we use, agent and environment. The learning system is called an agent. This system observes the environment and performs some actions. Based on the action, this system gets some reward. If the agent performs some bad actions, the system gets some penalties or can call it a negative reward. By getting positive and negative rewards, the agent learns from the environment. The agent must learn by itself what is the best strategy, called a policy so that the agent gains positive rewards over time. A policy defines what action the agent should choose when it is in a given situation.
For example, We all heard that the chess champion is defeated by the robot. Here the robot is the agent. Chess has thousands of moves and simply can’t defeat chess champions. You need the information about lots of moves. Here the robot learns from the thousands of moves itself. Each bad move by the robot in chess gives a negative reward and each good move gives a positive reward. The robot learns from this and gets better and better over time.
Thank you.
