# Applications of Matrices and Vectors

Mathematics I · Matrices and Vectors · https://tryals.app/learn/mathematics-i/applications-of-matrices-and-vectors

## Geometric & System Applications

Linear algebra unites diverse problems into one framework: represent a problem with **vectors and matrices**, and computation becomes **matrix arithmetic**.

**Geometric transformations**: A matrix is a linear map. Rotating, scaling, reflecting, or shearing points uses a **matrix–vector product** $\mathbf{x} \mapsto A\mathbf{x}$. Chaining them is **matrix multiplication**, and the determinant measures area scaling and orientation flips.

**Solving real systems**: Chemical equations, electrical circuits, and economic models reduce to a linear system $A\mathbf{x} = \mathbf{b}$, solved via Gaussian elimination or $\mathbf{x} = A^{-1}\mathbf{b}$.

## Networks, Data & Pitfalls

**Networks and data**: A network's connections form an **adjacency matrix** where powers count path lengths. A dataset is a matrix of samples and features where matrix computations handle projections and correlations.

| Application | Core Operation | Purpose |
| --- | --- | --- |
| **Graphics** | $A\mathbf{x}$ & multiplication | Transformations |
| **Circuits** | $A\mathbf{x} = \mathbf{b}$ | Solving systems |
| **Networks** | Matrix powers | Counting paths |

> **Common pitfall**: Viewing applications as unrelated tricks. Graphics, circuits, and networks are the same core operations applied to different data.

## Practice questions

9 of this lesson's 13 practice questions, with answers. The full set is in the app.

### 1. Writing a system as Ax = b lets you solve it as x = A⁻¹b - but numerical software almost never computes the inverse. Why is the formula that makes the theory clear the wrong way to compute the answer?

A. Because x = A⁻¹b gives a different answer from elimination, and elimination is the correct one.
B. Because forming A⁻¹ does more work and adds more rounding error.
C. Because A⁻¹ does not exist for most matrices arising in practice, so the formula rarely applies.
D. Because computing an inverse requires the determinant, which cannot be calculated for large matrices.

**Answer:** B. Because forming A⁻¹ does more work and adds more rounding error.

**Why:** The formula says what the solution *is* and is not an instruction for finding it: computing A⁻¹ means solving n systems, one per column of the identity, where elimination solves the one you care about, roughly n³ against n³/3. Cramer's rule is another exact formula that is catastrophic to compute with. The clean expression and the good algorithm answer different questions.

Page: https://tryals.app/practice/mathematics-i/applications-of-matrices-and-vectors/writing-a-system-as-ax-b-lets-you-solve-it-as-x-a-b-but

### 2. Order the stages of attacking a real problem with linear algebra.

**Answer:**

1. Identify the unknown quantities and collect them into a vector x
2. Express each condition of the problem as a linear equation
3. Assemble the coefficients into a matrix A and the constants into a vector b
4. Row-reduce the augmented matrix
5. Read off the solution, or conclude that there is none

**Why:** Name the unknowns, write the conditions, package them as A and b, reduce, and interpret. The last step matters: the answer may be a unique solution, a family with free variables, or no solution at all, and the reduced matrix tells you which.

Page: https://tryals.app/practice/mathematics-i/applications-of-matrices-and-vectors/order-the-stages-of-attacking-a-real-problem-with-linear-algebra

### 3. Match each operation to the job it does.

**Answer:**

- Matrix-vector product → Applies one transformation to one point
- Matrix product → Composes two transformations into one
- Determinant → Gives the area scaling factor of a transformation
- Adjacency matrix → Encodes which nodes of a network are connected

**Why:** These four cover most of the unit's applications: transform a point, chain transformations, measure the distortion, and encode a network. Every application in the lesson is one of these used on different data, the abstraction-and-reuse message of the whole unit.

Page: https://tryals.app/practice/mathematics-i/applications-of-matrices-and-vectors/match-each-operation-to-the-job-it-does

### 4. A three-node network has adjacency matrix with rows (0, 1, 1), (1, 0, 1) and (1, 1, 0). The entry in row 1, column 1 of the square of that matrix counts the walks of length 2 from node 1 back to itself. Compute it.

**Answer:** 2

**Why:** The entry is 0(0) + 1(1) + 1(1) = 2: you can go out to node 2 and back, or out to node 3 and back. Powers of an adjacency matrix count walks of that length, which turns a graph question into plain matrix multiplication.

Page: https://tryals.app/practice/mathematics-i/applications-of-matrices-and-vectors/a-three-node-network-has-adjacency-matrix-with-rows-0-1-1-1-0

### 5. The matrix with rows (2, 0) and (0, 2) is applied to the plane. By what factor does it multiply areas?

**Answer:** 4

**Why:** The determinant is 4, so areas quadruple, a unit square becomes a 2 by 2 square. Doubling every length doubles the area twice over, which is why the scaling factors multiply rather than add.

Page: https://tryals.app/practice/mathematics-i/applications-of-matrices-and-vectors/the-matrix-with-rows-2-0-and-0-2-is-applied-to-the-plane-by

### 6. Rotation by 90 degrees is the matrix with rows (0, -1) and (1, 0). Apply it to the point (3, 1). What is the first component of the image?

**Answer:** -1

**Why:** The first component is 0(3) + (-1)(1) = -1, and the full image is (-1, 3). The point has swung a quarter turn counterclockwise, exactly as a rotation should. Applying a transformation is nothing more than a matrix-vector product.

Page: https://tryals.app/practice/mathematics-i/applications-of-matrices-and-vectors/rotation-by-90-degrees-is-the-matrix-with-rows-0-1-and-1-0

### 7. Sort each object by how it is naturally represented.

**Answer:**

- Represented by a vector: A point in the plane, The direction of a line in space, A single displacement in space
- Represented by a matrix: A rotation of the plane, A dataset with rows as samples and columns as features, The connections of a network

**Why:** Points, directions and displacements are single lists, vectors. Transformations, datasets and network connections all relate rows to columns and need two indices, matrices. Recognising which one a problem needs is the first step in the whole framework.

Page: https://tryals.app/practice/mathematics-i/applications-of-matrices-and-vectors/sort-each-object-by-how-it-is-naturally-represented

### 8. Complete the interpretation of adjacency matrix powers.

**Answer:** Raising an adjacency matrix to the power k counts **the walks of length k between each pair of nodes**.

**Why:** Every multiplication by the adjacency matrix extends each walk by one edge, so the k-th power counts walks of length k between each pair of nodes. The row-times-column rule is exactly a sum over all possible intermediate nodes, which is why the rule is the right one for composition.

Page: https://tryals.app/practice/mathematics-i/applications-of-matrices-and-vectors/complete-the-interpretation-of-adjacency-matrix-powers

### 9. Solving Ax = b by computing A⁻¹ then multiplying is mathematically correct and numerically discouraged. What is the objection?

A. That matrix multiplication cannot be performed accurately in floating point
B. That A⁻¹ may not exist, whereas elimination always produces an answer
C. That the two methods give different answers, and elimination's is correct
D. Both cost and accuracy: n systems instead of one, with more error

**Answer:** D. Both cost and accuracy: n systems instead of one, with more error

**Why:** The inverse costs n systems where you needed one, roughly three times the work, and every extra operation rounds. Both routes are exact on paper and only one is exact in floating point, which is why `numpy.linalg.solve` exists and calling `inv` first is the classic beginner's tell.

Page: https://tryals.app/practice/mathematics-i/applications-of-matrices-and-vectors/solving-ax-b-by-computing-a-then-multiplying-is-mathematically