![](https://crypto4nerd.com/wp-content/uploads/2023/06/0dkia2_REr9tHg_We.png)
Nowadays, Chat Gpt is a hot topic in the computer world. Everyone is talking about the pros and cons of the Chat Gpt. Whether it has more pros than cons is another topic but it has started a hype about a branch of computer science called “Artificial Intelligence”. Even non-tech people are taking an interest in this field. Artificial Intelligence is not a new dimension of technology. This term was coined about 70 years ago. Artificial intelligence is a broad sector of technology and we can think of machine learning as a subset of it.
Machine learning is a branch of Artificial intelligence that focuses on training data and building an algorithm to predict test data. The algorithm learns from training examples and slowly tries to improve its accuracy. It’s something similar to how a human learns. It is widely used in various fields like medicine, email filtering, speech recognition, agriculture etc where we can’t predict on the basis of conventional algorithms as the data may be so much diverse that it can’t be fitted into the existing algorithms.
Machine learning has 3 types.
(1) Supervised learning
(2) Unsupervised learning
(3) Reinforcement learning
Supervised learning mainly deals with labeled data. Here we have some training data which we fit into the algorithm and then we try to predict for testing data. The prediction data can be continuous or discrete, it depends on the type of data.
For example: let’s say you are a professional gambler and you bet on cricket matches. Then before making any bet, you will analyze some features like the condition of the pitch, weather, injured players etc. After all these analyses you are going to place a bet on a team. Here the prediction will be binary either the team you choose will win or lose but you have some labeled features, based on what you are trying to make a prediction. let’s consider another classic example of supervised learning where a broker has to sell a house so he decides to make a list of some features on which the price of a house depends like no. of bedrooms, living area etc. Based on these features the pricing of a house is made. Here in both examples, we have labeled data and based on those labeled data we have to make a prediction.
Some algorithms that are used in supervised learning are Linear Regression, Logistic Regression, Support Vector Machine, Decision Tree Classifier, Random Forest Classifier, XGBoost etc.
Unsupervised learning deals with unlabelled data. Here we have some unlabelled data, in which we have to find some features that will affect our interest. Here our model learns patterns and relationships from unlabelled data. Unlike supervised learning, there are not any target values on which our model will train itself. Unless our model will explore the data and try to find some hidden features and patterns.
It is used by a lot of commercial companies where companies want to understand their customer base to tailor their marketing strategies. They have a very large dataset containing information about their customers like age, gender, browsing behavior, purchase history etc. But they do not have any predefined labels on which they can classify their customers so that they can focus on increasing their sales. In this scenario, unsupervised learning can help to segment the customer base based on their shared characteristics or features. Then they can make marketing campaigns focused on the targeted segment’s preferences and needs. For example, one segment consists of tech-savvy customers who are interested in the latest gadgets and another segment may be interested in traditional products. This is just one example of unsupervised learning that can be applied to various domains like finance, health industry, fraud detection, image analysis etc.
Some algorithms that are used in unsupervised learning are K-means clustering, Hierarchical clustering, DBSCAN(Density-Based Spatial Clustering of Applications with Noise), Principal Component Analysis(PCA), t-SNE(t-Distributed Stochastic Neighbor Embedding), Gaussian Mixture Models(GMM), Anomaly detection algorithms etc.
Reinforcement learning focuses on training the agent on sequential decisions in an environment to maximize a cumulative reward. It is based on the way how humans learn by interacting with the environment. The agent learns through a trial and error process, exploring different actions and observing their outcomes to improve the decision-making ability.
Agent: An entity that can learn and take decisions or actions in an environment. Generally, an agent would be algorithms, robots or any system capable of learning and taking decisions or actions.
Environment: It is an external world or simulation in which the agent operates. It can be a physical world, a virtual world or a software simulation.
Reward: It is a numerical signal provided to the agent for the action taken by the agent. It can be positive, negative or zero depending on the desirability of actions.
Reinforcement learning employs different algorithms to learn and improve the cumulative reward. Some algorithms are Q-learning, SARSA, Deep Q-Networks, Proximal Policy Optimization, Actor-Critic methods etc.
Reinforcement learning has found applications in various domains like robots, Autonomous Driving, Financial Trading, Energy Management etc.
Here the blog ends, In the next blog we will deep dive into supervised learning algorithms. We will look into the Linear Regression algorithm.