Three main categories of machine learning with examples of usage.
Machine Learning is a subset of artificial intelligence that primarily focused on their experience and making predictions based on it. The main concept behind this is the study of how to build applications that exhibit this iterative improvement.
There are many ways to frame this idea, but mainly there are three major recognized categories.
1. Supervised Learning
Supervised learning is one of the most basic types of machine learning and it is trained on labeled data. At the end of the training, the algorithm then finds relationships between the parameters given. supervised learning is extremely powerful when used in the right circumstances.
It’s a learning mechanism that the model is trained with an already available correct data set. We start with a training data set that is already labeled inputs and outputs. The model is specifically knowing what to identify. And the model is training until it can identify similar data that we feed. Use to detect patterns and similarities in the input data set. Human interference is at a considerable level in this pattern.
There are mainly two types of supervised learning problems namely classification and regression. In classification, data get classified into different labels. In regression, it represents the prediction of continuous quantity. An example of classification is the labeling of emails as spam or non-spam. A regression example is the prediction of humidity with the change of temperature. Linear regression, Logistic regression, K Nearest Neighbour, Random Forest are examples of algorithms used in supervised learning. Supervised learning is widely used for risk evaluation and forecasting sales.
Face Recognition: Facebook uses a supervised learning algorithm to recognize faces and delete fabricated accounts. Having a system that takes a photo, searches face, and imagines who is in the photo (tagging) is a supervised process
Spam Classification: Modern email systems encountered a spam filter. That spam filter is a supervised learning system. These systems are used to filter out malicious emails because that the user is not harassed by them and provide and labels (spam/non-spam) from the email system.
2. Unsupervised Learning
Unsupervised learning is very much the opposite of supervised learning. It infers patterns from a dataset without reference to known or labeled, outcomes and fed a lot of data, and given the tools to understand the properties of the data
We feed unlabeled or untagged row data as input and getting meaningful output data. This technique is similar to how the human mind works to categorize things thus it’s called artificial general intelligence. Human interference is minimal. the model must learn to map the data set before the observations.
There are mainly two types of unsupervised learning problems namely clustering and association. In clustering data get grouped by the usage patterns of the user for targeted marketing. In association, it identifies hidden patterns among data and uses them for generating predictions. An example of clustering is identifying the internet and cellular usage of the people by a certain telecommunication provider and grouping them into different priority types and providing each of them with the best promotions.
An example of association is the customer product buying patterns recognition in supermarkets where frequently both together items are identified and kept closed in the shelves so customers can easily put them in the cart without going through different shelves. K-Means, C-Means are examples of algorithms used in unsupervised learning. Unsupervised learning is widely used for recommendation systems and detecting anomalies.
Grouping User Logs: Unsupervised learning holds the advantage of being able to group user logs and issues. This can help businesses to identify key issues to customers and correct them by improving a product or developing a FAQ to deal with common issues.
Recommender Systems: A video recommendation system has been met by YouTube or Netflix. These systems are often placed in the unattended field. Perhaps we only know its duration, its genre, etc. The recommendation system can see the relationship in the data that we previously watched and promptly suggested.
3. Reinforcement learning
Reinforcement learning is inspired directly by how people learn in their lives from data. It features an algorithm that improves upon itself and learns from new situations using a trial-and-error method.
“ Without using a predefined dataset a model that learns from mistakes”
is the simplest term that can be given. This means we expose the model to an unknown environment and let it do things. for the positive things it does we give positive points and mistakes it does we add negative points. This model has been much more useful in many physical activities.
There are mainly three types of reinforcement problems classification namely policy-based, value-based, and model-based. A policy-based approach, observes the behaviors of the agent in a certain environment. In value-based approach models, it observes different actions taken by an agent in a certain environment and whether they are good or bad. In a model-based approach, it observes how an agent learns from in a certain environment. Q-Learning, Deep Q-learning, Policy gradient are examples of algorithms used in reinforcement learning. Reinforcement learning is widely used in training agents for playing games.
Video Games: One of the most common places to look at reinforcement learning is in learning to play games. Look at Google’s reinforcement learning applications, Alpha Zero, and AlphaGo which learned to play the game Go. And Mario is a common example.
Resource Management: Reinforcement learning is a good way to navigate complex environments It can deal with the necessity to balance certain demands. Google’s data centers, used reinforcement learning to balance the need to satisfy power requirements.
