Linear Algebra Language in the AI Age

How Vectors and Matrices Describe the World

Author

He Wang

Published

June 6, 2026

Welcome

NoteLinear Algebra Language in the AI Age

A visual, story-driven first introduction to vectors, matrices, data, and AI.

This book teaches linear algebra as a language for describing the world: how information is represented, transformed, compared, approximated, compressed, and learned from.

Most people first meet linear algebra as a page full of symbols.

Rows of numbers.
Rules for matrix multiplication.
Systems of equations.
Determinants.
Eigenvalues.

Those tools matter. But they are not the beginning of the story.

The beginning is much simpler:

The world produces information, and linear algebra gives us a language for organizing it.

A vector can describe a person, an image, a document, a song, a gene, a customer, a sensor reading, or a word. A matrix can rotate a shape, solve a system, filter an image, rank webpages, compress data, or act as one layer of a neural network. A basis can be a coordinate system, a dictionary, or a language for expressing the same object in different ways.

This book is built around one guiding idea:

Linear algebra is not only a collection of formulas.
It is a way of seeing.

The Story in One Page

Linear algebra begins with numbers that have meaning.

A list of numbers becomes a vector.
A vector becomes a point in data space.
Many vectors become a dataset.
A matrix becomes a machine that moves, changes, filters, combines, or compresses those vectors.

Then geometry enters.

Length tells us how large something is.
Distance tells us how far apart two objects are.
Angle tells us whether two objects point in similar directions.
Orthogonality tells us when two pieces of information are independent.
Projection tells us the best shadow of one object on another.

Then hidden structure appears.

Eigenvectors reveal directions that survive a transformation.
Eigenvalues measure how strongly those directions stretch or shrink.
The singular value decomposition reveals the most important patterns inside a matrix.
Fourier and Haar ideas show how signals and images can be built from simpler pieces.
Modern AI uses these same ideas to represent images, text, recommendations, and neural networks.

The story moves from ordinary numbers to the mathematical grammar of data and intelligence.

Who This Book Is For

This book is written for readers who want to understand linear algebra before they are buried under notation.

It is especially suitable for:

  • first-year college students learning linear algebra for the first time, (or high school students with curiosity about data and AI,)
  • data science and machine learning beginners,
  • students in biology, business, engineering, social science, and computer science,
  • teachers who want a more visual and story-driven path into the subject,
  • general readers who want to understand the mathematics behind modern technology.

No prior linear algebra is assumed. The reader should be comfortable with algebra, graphs, and basic functions. Python is used as a tool for exploration, but the central ideas are explained before computation.

A follow-up book on advanced linear algebra is available: Linear Algebra in the AI Age: Geometry, Computation, and Data.

What Makes This Book Different

Many linear algebra books begin with definitions and procedures. This book begins with meaning.

Each chapter tries to answer three questions:

  1. What problem is this idea trying to solve?
  2. What does the idea look like geometrically or visually?
  3. Where does the idea appear in data, computation, or AI?

The style is:

Feature What it means in this book
Concept-first Ideas are introduced through meaning before formal notation.
Visual-first Geometry, pictures, and intuition guide the algebra.
Story-driven Each topic grows naturally from the previous one.
Computation-supported Python helps readers experiment and see ideas in action.
AI-era friendly Examples connect vectors and matrices to images, text, recommendation systems, and neural networks.

The guiding principle is:

Do not introduce a formula before the reader understands the question the formula is answering.

The Six Big Ideas

The entire book is organized around six recurring ideas.

1. Representation

Before we can compute with something, we must represent it.

A house, word, image, user, product, medical record, or signal can become a vector. Linear algebra begins when meaningful objects become numerical objects.

2. Transformation

A matrix is a machine for changing vectors.

It can stretch, rotate, reflect, shear, project, filter, combine, compress, rank, and learn.

3. Similarity

The dot product, length, distance, angle, and cosine similarity help us compare objects.

This is the beginning of search, clustering, recommendation systems, and semantic comparison.

4. Approximation

Real data is noisy, incomplete, or inconsistent.

Projection and least squares help us find the best possible answer when an exact answer does not exist.

5. Compression

Many complicated objects contain simpler structure.

Linear algebra helps us keep the important part and remove the less important part.

6. Hidden Structure

Eigenvectors, eigenvalues, singular values, and singular vectors reveal patterns that are not obvious from the raw data.

These ideas are the foundation of PCA, ranking, stability, denoising, image compression, and many machine learning methods.

Roadmap of the Book

Part I. Seeing Data as Geometry

This part introduces numbers, vectors, combinations, and data points.

Main question:

How can objects in the world become points in space?

Suggested chapters:

Part II. Matrices as Machines

This part treats matrices as actions, not just tables.

Main question:

How does information change?

Suggested chapters:

Part III. The Geometry of Similarity and Approximation

This part develops the geometry of length, angle, projection, and orthogonality.

Main question:

How do we compare and approximate information?

Suggested chapters:

Part IV. Hidden Directions

This part studies eigenvectors, stability, ranking, and energy.

Main question:

What directions reveal the structure of a transformation?

Suggested chapters:

Part V. The Matrix Microscope

This part introduces SVD, compression, PCA, Fourier ideas, and Haar ideas.

Main question:

How can simple structure be found inside complex data?

Suggested chapters:

Part VI. Linear Algebra and AI

This part connects the language of linear algebra to images, text, neural networks, recommendation systems, and modern AI.

Main question:

Why is linear algebra one of the main languages of artificial intelligence?

Suggested chapters:

How to Use This Book

Read slowly. Draw pictures. Try small examples by hand. Run the Python code when it appears. Ask what each symbol means in the story.

When you see a vector, ask:

What does this vector represent?

When you see a matrix, ask:

What does this matrix do?

When you see a dot product, ask:

What kind of similarity is being measured?

When you see a projection, ask:

What is being approximated?

When you see a decomposition, ask:

What structure is being revealed?

Use AI as a companion for explanation, examples, and reflection, but not as a replacement for thinking. The goal is not to finish pages quickly. The goal is to build a new way of seeing.

Start Reading

Begin here:

Introduction and Overview