Supervised vs Unsupervised vs Semi-supervised vs Reinforcement Learning
Machine learning systems are grouped by how much supervision they get: labelled data (supervised), no labels (unsupervised), a few labels (semi-supervised), or rewards (reinforcement).
Supervised learning: the training set includes the desired solutions, called labels. Classification (is this email spam?) and regression (predict a house price) are the typical tasks. Key algorithms: linear and logistic regression, k-nearest neighbours, SVMs, decision trees, random forests and neural networks.
Unsupervised learning: the data has no labels, so the system tries to find structure on its own. Typical tasks are clustering, visualization, dimensionality reduction and anomaly detection. Key algorithms: k-means, hierarchical clustering, PCA.
Semi-supervised learning mixes a lot of unlabelled data with a little labelled data (Google Photos recognises the same person across photos, then you name them once). Reinforcement learning has an agent that acts in an environment, gets rewards or penalties, and learns a policy, the best strategy to maximise reward (robots learning to walk, game bots).
Key points
- Supervised = labelled data, a teacher exists.
- Unsupervised = unlabelled data, the system finds groups or structure.
- Semi-supervised = a little labelled plus a lot unlabelled.
- Reinforcement = agent, environment, reward, and a learned policy.
- Clustering and dimensionality reduction are unsupervised; classification and regression are supervised.
Worked example
Sorting emails into spam or not spam with labelled examples is supervised. Grouping customers into segments with no predefined groups is unsupervised. A bot that learns chess by playing and getting win or lose signals is reinforcement learning.
Common questions
Is clustering supervised?
No, clustering is unsupervised because no labels are given.
What is a policy in reinforcement learning?
The strategy the agent learns that tells it which action to take in each situation to get the most reward.
Practice questions on this topic
Try these, then open the app to check your answers with explanations.
- What is The Most Important Supervised Learning Algorithms?
- A. k-Nearest Neighbors.
- B. Linear Regression.
- C. Logistic Regression.
- D. All of them are true
- what is the Unsupervised Learning Algorithms used in clustring ?
- A. K-Means
- B. DBSCAN
- C. HCA
- D. All of them are true
- What is the main characteristic of semi-supervised learning?
- A. The training set includes desired solutions
- B. The training data is unlabeled
- C. It often has plenty of unlabeled instances and few labeled instances
- D. The agent can observe the environment
- In Unsupervised Learning Algorithms used to Visualization and dimensionality reduction
- A. PCA
- B. One-class SVM
- C. Apriori
- D. All of them are true
- Which type of learning relies on a teacher or desired solutions?
- A. Supervised Learning
- B. Unsupervised Learning
- C. Semi-supervised Learning
- D. Reinforcement Learning
- What is the main characteristic of supervised learning?
- A. The training set includes desired solutions
- B. The training data is unlabeled
- C. It often has plenty of unlabeled instances
- D. The agent can observe the environment
Related topics
- What is Machine Learning?
- Overfitting and Underfitting in Machine Learning
- Batch Learning vs Online Learning
- Confusion Matrix, Precision, Recall and F1 Score
- ROC Curve and AUC Explained
- K-Fold and Stratified K-Fold Cross-Validation
- All topics: Machine Learning
BASMAGLY is an independent student project, not an official Delta University site. Check lectures and your instructor for the final word.