Supervised Learning and Neural Networks#

Stage 1 · 🧠 Introduction to Deep Learning · Lesson 02 of 17 · beginner

◀ Previous · What is a Neural Network? · Next · Why Deep Learning is Taking Off ▶ · ↑ Section

Important

✨ AI-generated content. This page was written with the assistance of an AI language model and is provided as a learning aid. Despite careful review, it may still contain mistakes, omissions, or out-of-date information. Whether you are new to the topic, a team lead, or a senior practitioner, treat it as a starting point rather than an authoritative reference: read it critically and independently verify anything you act on (code, commands, figures, and factual claims) against official documentation and primary sources before relying on it.

Learning a mapping#

Nearly all the commercial value of neural networks comes from supervised learning: you are given labelled examples — inputs \(x\) paired with correct outputs \(y\) — and you learn the function that maps \(x \to y\). House features → price, an ad and a user → click-or-not, an image → the object it contains: each is the same problem, learn \(x \to y\) from examples.

Matching the architecture#

The input type dictates the network. A standard (fully connected) network suits tabular features (home prices, ad click-through). Convolutional networks (CNNs) suit images — computer vision, photo tagging. Recurrent networks (RNNs) suit sequences — audio → text, language translation. Rich inputs (autonomous driving from image + radar) often use hybrid / custom designs. Same learning principle, different wiring.

Structured vs unstructured#

Data splits into two kinds. Structured data is table-like: rows and columns where each feature has a clear meaning (a house’s size, a user’s age). Unstructured data is raw audio, images, text, where the features are pixels or waveform samples or words and no single feature means much on its own. Classical algorithms handle structured data well but struggle on unstructured data.

Why it matters#

This is exactly where neural networks changed the game: they made computers far better at unstructured data — speech, vision, language — unlocking applications that were out of reach a decade ago. Much of the economic value, meanwhile, still comes from structured data inside companies, where accurate predictions on large databases translate directly into money.

Hint