Notes – Types of Machine Learning
Machine Learning (ML) is categorized into different types based on how a model learns from data. The three main types of ML are:
- Supervised Learning – Learning with labeled data.
- Unsupervised Learning – Finding patterns in unlabeled data.
- Reinforcement Learning – Learning through rewards and penalties.
Each type has different use cases and applications.
1. Supervised Learning
Supervised learning involves training a model using labeled data, where input-output pairs are provided. The model learns to map inputs to correct outputs.
Key Features:
- Requires labeled data.
- Used for classification and regression tasks.
- Learns from past data and makes future predictions.
Examples:
- Spam Detection – Classifying emails as spam or not spam.
- Credit Score Prediction – Predicting loan approval chances based on financial history.
Common Algorithms:
| Algorithm | Description |
|---|---|
| Linear Regression | Predicts continuous values (e.g., house prices). |
| Logistic Regression | Used for binary classification (e.g., spam or not spam). |
| Decision Trees | Splits data into branches for decision-making. |
| Support Vector Machines (SVM) | Finds the best decision boundary for classification. |
2. Unsupervised Learning
Unsupervised learning works with unlabeled data, meaning the model identifies patterns and structures on its own.
Key Features:
- No labeled output; the model finds hidden patterns.
- Commonly used for clustering and dimensionality reduction.
- Helps in exploratory data analysis.
Examples:
- Customer Segmentation – Grouping customers based on purchasing behavior.
- Anomaly Detection – Identifying fraudulent transactions in banking.
Common Algorithms:
| Algorithm | Description |
|---|---|
| K-Means Clustering | Groups similar data points into clusters. |
| Hierarchical Clustering | Creates a tree-like structure of clusters. |
| Principal Component Analysis (PCA) | Reduces data dimensions while retaining important features. |
3. Reinforcement Learning (RL)
Reinforcement Learning (RL) is based on the concept of learning through rewards and penalties. An agent interacts with an environment and takes actions to maximize rewards.
Key Features:
- Involves an agent, actions, and an environment.
- No predefined correct answers; learning is based on trial and error.
- Used in decision-making and automation tasks.
Examples:
- Game AI – AI playing chess and improving its strategy.
- Self-Driving Cars – Learning to navigate roads by trial and error.
Common Algorithms:
| Algorithm | Description |
|---|---|
| Q-Learning | Uses a Q-table to store action-reward values. |
| Deep Q-Networks (DQN) | Uses neural networks to approximate Q-values. |
| Policy Gradient Methods | Optimizes policies directly for better decisions. |
Comparison of ML Types
| Feature | Supervised Learning | Unsupervised Learning | Reinforcement Learning |
|---|---|---|---|
| Data Type | Labeled | Unlabeled | Feedback-based |
| Goal | Predict outcomes | Find patterns | Maximize rewards |
| Common Uses | Classification, Regression | Clustering, Anomaly Detection | Robotics, Game AI |
| Example | Email spam detection | Customer segmentation | Self-driving cars |
