Machine learning sounds complex, but it need not be. At its core, it is just about teaching computers to learn patterns from data. Irrespective of your background, this series breaks down AI/ML concepts with simple examples and real-world use cases.
Welcome to the series!
In the first post, we will explore the three types of machine learning – Supervised, Unsupervised, and Reinforcement learning.
Supervised Learning
In Supervised learning, the model learns from a labeled data to predict outcomes on new, unseen data. Here labeled data refers to the input and the known output.
labeled data = input + known output
Sample Python Code Snippet:
from sklearn.linear_model import LinearRegression
# Predict house prices based on square footage
model = LinearRegression()
model.fit(X_train, y_train) # where y_train = known prices
Real-world use cases:
Predicting customer churn, email classification, credit scoring, whether a loan will default.
Unsupervised Learning
In Unsupervised learning, the model explores unlabeled data to find hidden patterns or groupings.
Sample Python Code Snippet:
from sklearn.cluster import KMeans
# Grouping customers based on behavior
kmeans = KMeans(n_clusters=3)
kmeans.fit(customer_data)Real-world use cases:
Customer segmentation, anomaly detection, organizing news articles or image compression.
Reinforcement Learning
The model (usually referred as an agent) learns by interacting with an environment, receiving rewards or penalties for its actions. You can think this as like trial and error.
In a real-world analogy, this is like teaching a dog tricks with treats – the dog learns what gets rewarded.
Real-world use cases:
Self-driving cars, robotics, game playing (e.g., AlphaGo) or recommendation systems that adapt over time.
Let’s put above in a simple summary as below:
| Learning Type | Learns From | Key Output | Example Use Case |
| Supervised Learning | Labeled data | Predictions | Predicting churn |
| Unsupervised Learning | Unlabeled data | Patterns/Groups | Customer Segmentation |
| Reinforcement Learning | Rewards & actions | Optimal Strategies | Self-driving cars |
Coming Next
In Part 2, We will delve into two powerful prediction models: Classification and Regression. Also, explanation on when a model “overfits” or “underfits”.