3.1.1. Types of AI Algorithms

In supervised learning, the algorithm is trained on labeled data, where the desired output is known. The algorithm learns to map input data to the correct output, minimizing the error between predicted and actual outputs. Common algorithms in this category include:

• Linear Regression
• Support Vector Machines (SVM)
• Decision Trees
• Neural Networks

Supervised learning is a type of machine learning where an algorithm is trained using a dataset that contains the inputs and the corresponding correct outputs, known as labeled data. The learning process involves the algorithm adjusting its internal parameters to minimize the difference, or error, between the predicted outputs and the actual outputs. This type of learning is used in many applications where the outcomes are known and can be used to teach the system. For example, in a simple linear regression model, the algorithm learns the relationship between variables, allowing it to make predictions based on new input data.

Imagine a teacher (the algorithm) training students (the model) with labeled examples—like providing math problems (input data) along with the correct answers (desired output). Just as students learn to solve similar problems by practicing with given solutions, the algorithm learns from past examples to make accurate predictions on new, unseen data.

Unsupervised learning algorithms are used to find patterns or structure in data that is not labeled. The goal is to identify the underlying structure or relationships within the data, such as clustering similar data points. Common algorithms include:

• K-Means Clustering
• Principal Component Analysis (PCA)
• Generative Adversarial Networks (GANs)

Unsupervised learning is a category of machine learning where the algorithms work with data that does not have labels or predefined outcomes. The primary objective is to analyze the input data and find hidden patterns or groupings. For instance, in K-Means clustering, the algorithm groups similar data points together based on their characteristics without prior knowledge of specific labels. This type of learning is useful in scenarios where you want to explore data to see what structures might exist.

Think of unsupervised learning like sorting a box of assorted candies without knowing the types beforehand. As you begin to group similar candies (based on color, shape, or size), you may identify clusters without ever having seen a label indicating what each candy is. This process reveals an underlying structure in the data.

In reinforcement learning, an agent learns by interacting with an environment and receiving feedback through rewards or punishments. The agent's goal is to maximize the cumulative reward over time by exploring and exploiting the environment. Popular algorithms include:

• Q-Learning
• Deep Q-Networks (DQN)
• Policy Gradient Methods

Reinforcement learning is a learning paradigm where an agent learns to make decisions by performing actions in an environment and receiving feedback in the form of rewards or penalties. The agent seeks to optimize its actions to achieve the maximum cumulative reward. For example, in a game setting, the agent learns which actions lead to wins, maintaining a balance between trying new strategies (exploration) and using known strategies that yield rewards (exploitation). This feedback allows the agent to improve its decision-making over time.

Consider a dog learning tricks. Each time it successfully performs a trick, it receives a treat (reward). If it fails, it might face a negative consequence, like no treat at all. Over time, the dog learns which tricks yield the best outcomes, just as a reinforcement learning agent learns the best actions to take in its environment.

--