Machine learning can be grouped into three main categories based on how models learn from data.
🎯 1. Supervised Learning
In supervised learning, the model learns from labeled data.
- Input: Features (e.g. image, sentence, numbers)
- Output: Label (e.g. cat or dog, sentiment)
The goal:
Learn to predict the correct label from input.
Examples:
- Spam detection
- Sentiment analysis
- Image classification
🔍 2. Unsupervised Learning
In unsupervised learning, the model receives unlabeled data and tries to find patterns or structure.
There are no “correct answers” — the model groups or compresses data on its own.
Examples:
- Clustering customers by behavior
- Dimensionality reduction (e.g. PCA)
- Anomaly detection
🎮 3. Reinforcement Learning (RL)
In reinforcement learning, the model (called an agent) learns by interacting with an environment.
It receives:
- Rewards for good actions
- Penalties for bad ones
Over time, it learns a policy:
“What action should I take in each situation to maximize reward?”
Examples:
- Game-playing AIs (e.g. AlphaGo)
- Robotics
- Autonomous navigation
📊 Comparison Table
| Type | Data | Output | Example | |--------------------|--------------|---------------|------------------------------| | Supervised | Labeled | Prediction | Spam filter, price prediction| | Unsupervised | Unlabeled | Pattern/group | Customer segmentation | | Reinforcement | Dynamic env. | Actions | Game bots, self-driving cars|
🧠 Summary
- Supervised: Learn from examples with answers.
- Unsupervised: Discover hidden structure.
- Reinforcement: Learn by trial, error, and reward.
✅ Self-Check
- Can you give an example of each learning type?
- Why is reinforcement learning different from the other two?
- Which type would you use to detect fraud?