Supervised Learning:

1. How it Learns:

• Like teaching with examples. You show the machine lots of labeled examples (like photos labeled as cats or dogs) so it can learn from them. In simple language a trainer train the ML model by providing input and output so they can perform next action on the basis of the that input and output.

2. Use Case:

• Predicting things based on what it’s learned before. For instance, guessing if a new picture is a cat or a dog.

• Predictive modeling, classification, regression

3. Algorithms:

• Linear Regression: Predicts continuous values (e.g., housing prices).

• Logistic Regression: Classifies data into categories (e.g., spam or not spam emails).

• Support Vector Machines (SVM): Effective for both classification and regression tasks.

• Decision Trees and Random Forest: Used for classification and regression, handling complex datasets.

• Naive Bayes: Great for text classification and spam filtering.

Unsupervised Learning:

1. How it Learns:

• It’s like sorting things without any instructions. The machine looks for patterns or groups in the data all by itself. There is no no trainer who train the model . On the basis of input they grouping the data (clustering)

2. Use Case:

• Grouping similar things together without being told what makes them similar. For example, sorting different colored marbles by their shades.

3. Algorithms:

• K-Means Clustering: Segments data into clusters based on similarity.

• Hierarchical Clustering: Organizes data into a tree-like cluster hierarchy.

• Principal Component Analysis (PCA): Reduces dimensions in data while retaining important information.

• Apriori Algorithm: Discovers patterns in large datasets, often used in market basket analysis.

Reinforcement Learning:

1. How it Learns:

• Learning by doing and getting rewards or punishments. The machine tries different actions and learns from the consequences.

2. Use Case:

• Like teaching a robot to play a game by letting it try different strategies and rewarding good moves.

1. Algorithms:

2. Q-Learning: Teaches agents to take optimal actions in a specific environment.

3. Deep Q-Networks (DQN): Uses deep neural networks to tackle complex environments.

4. Policy Gradient Methods: Directly learn policies to maximize rewards.

5. Actor-Critic Models: Combines value estimation (critic) with action selection (actor) in an environment.