4.1.1 Linear Algebra Roadmap: Data as Vectors, Batches as Matrices

Linear algebra is the language AI uses to represent data and transform it. Do not start by memorizing proofs. First see what each object does in code.

Look at the Map First

The chapter flow is:

Idea	First meaning in AI
vector	one object written as numbers
matrix	many vectors stacked together, or a transformation
dot product	compare matching positions and add them up
matrix multiplication	many dot products at once
eigenvalue / eigenvector	important directions, useful for PCA intuition

Run the Smallest Loop

Create linear_algebra_first_loop.py and run it after installing numpy.

import numpy as np

student = np.array([90, 85, 92])
students = np.array(
    [
        [90, 85, 92],
        [70, 88, 75],
        [95, 91, 89],
    ]
)
weights = np.array([0.4, 0.2, 0.4])

single_score = student @ weights
all_scores = students @ weights

print("student_vector:", student)
print("matrix_shape:", students.shape)
print("single_score:", round(single_score, 2))
print("all_scores:", all_scores.round(2))

Expected output:

student_vector: [90 85 92]
matrix_shape: (3, 3)
single_score: 89.8
all_scores: [89.8 75.6 91.8]

If you accidentally use * instead of @, you get element-by-element multiplication, not a weighted score. This is the most useful early distinction.

Learn in This Order

Order	Read	What to focus on first
1	4.1.2 Vectors	object -> vector, length, dot product, cosine similarity
2	4.1.3 Matrices	batch data, matrix multiplication, X @ W + b
3	4.1.4 Eigenvalues and Eigenvectors	special directions, PCA intuition
4	4.1.5 Vector Spaces	basis, dimension, linear transformation

Evidence to Keep

Keep this page’s proof of learning as a small evidence card:

Math Object

vector, matrix, eigenvalue, basis, or vector space concept

Numeric Example

small numbers or NumPy snippet used to compute it

Visual Or Output

shape, transformed point, similarity score, eigen direction, or projection

Ai Link

where this appears in embeddings, batches, PCA, neural layers, or attention

Expected Output

calculation plus one sentence connecting it to an AI operation

Pass Check

You pass this roadmap when you can explain why one sample is a vector, why a batch is a matrix, what @ does, and why these ideas reappear in RAG similarity, PCA, and neural network layers.

Check reasoning and explanation

• The linear-algebra route is passed when you can read X @ W as both a shape operation and a batch of dot products.
• Keep one vector similarity example, one matrix transformation example, one PCA or eigenvector plot, and one SVD or rank check as evidence.
• The key explanation is not symbolic elegance; it is knowing what changed in direction, length, dimension, or redundancy.