I’ve been brushing up on my ML fundamentals lately—especially for interviews—and I decided to start documenting them visually and conceptually. This is the first one: a quick breakdown of “What is Machine Learning?” and the different types of ML.
What Even Is Machine Learning?
Arthur Samuel defined Machine Learning as the field that gives “computers the ability to learn without being explicitly programmed.”
But to me, this definition only clicks when you tie it to a task.
I like Tom Mitchell’s formalism better:
A computer program is said to learn from experience E, with respect to some task T, and some performance measure P, if its performance at the task improves with experience.
Let’s break that down with an example:
-
E → Watching how emails are labeled spam/not spam
-
T → Classifying new incoming emails
-
P → Accuracy (how many correct classifications)
If performance improves as it sees more labeled data—congrats, that’s learning.
Types of Machine Learning
Once you get the core idea, ML branches out into 3 primary types:
1. Supervised Learning
You're given labeled data—pairs of input and correct output—and the model learns from it.
-
Regression
Predict continuous values-
e.g., Linear Regression (straight line fit)
-
e.g., Logistic Regression (probability output)
-
-
Classification
Predict categories-
e.g., SVM draws a decision boundary between classes
-
Can be binary (0 or 1) or multi-class
-
2. Unsupervised Learning
No labels. You let the model find structure in the data.
-
e.g., Clustering using K-Means, KNN
-
Grouping similar items together without predefined categories.
-
3. Others
These are growing fast and are super relevant:
-
Reinforcement Learning (rewards/punishments, like game agents)
-
Recommender Systems (Netflix, YouTube suggestions)
This visual helped me anchor key ML ideas before diving into algorithms and model internals. It’s also a great snapshot to keep referring back to during interviews.
—Kavana
Comments
Post a Comment