Matrix operations — addition, subtraction (conditions)

Overview

Matrix addition and subtraction are element-wise operations that combine corresponding entries of matrices. The critical constraint: matrices must have identical dimensions — this isn't arbitrary algebra, it's grounded in the fact that each entry represents a specific position in a structured data arrangement.

[!intuition] Why Element-Wise?

Think of matrices as spreadsheets of data. If Matrix A tracks sales across 3 stores × 4 products, and Matrix B tracks returns for the same structure, adding them gives net sales. But you can't add a 3×4 sales sheet to a 2×5 inventory sheet — the positions don't correspond. Each matrix entry has semantic meaning from its row-column position; operations preserve this meaning by pairing entries that occupy the same logical slot.

Visual analogy: Stacking transparent grids — you can only add/subtract values that align perfectly. Mismatched grids leave some squares orphaned.

[!definition] Formal Definition

Let $A = [a_{ij}]_{m \times n}$ and $B = [b_{ij}]_{m \times n}$ be matrices of identical order $m \times n$.

Addition: $C = A + B = [c_{ij}]_{m \times n} \text{ where } c_{ij} = a_{ij} + b_{ij}$

Subtraction: $D = A - B = [d_{ij}]_{m \times n} \text{ where } d_{ij} = a_{ij} - b_{ij}$

Condition for both operations: Order of matrices must be identical ($m \times n$ for both).

[!formula] Derivation from First Principles

Why This Condition?

Start from the purpose: A matrix encodes relationships between row-items and column-items (e.g., students × subjects for grades).

1. Entry correspondence: $(i, j)$ in Matrix A refers to "row-entity $i$ paired with column-entity $j$"
2. Combining information: To add matrices, we need $a_{ij}$ and $b_{ij}$ to refer to the same relationship
3. Structural requirement: This forces both matrices to have the same $m$ (row count) and $n$ (column count)

What breaks if dimensions differ?

• $A$ is $2 \times 3$, $B$ is $2 \times 4$: The third column of $B$ has no partner in $A$
• $A$ is $3 \times 2$, $B$ is $\times 2$: The third row of $A$ has no partner in $B$

The operation becomes undefined — we'd need to invent values, which violates the deterministic nature of algebra.

Properties (Derived from Comutativity/Associativity of Real Numbers)

Since $c_{ij} = a_{ij} + b_{ij}$ and real number addition is commutative:

1. Commutativity: $A + B = B + A$

   • Why? $a_{ij} + b_{ij} = b_{ij} + a_{ij}$ for all $i, j$

2. Associativity: $(A + B) + C = A + (B + C)$

   • Why? $(a_{ij} + b_{ij}) + c_{ij} = a_{ij} + (b_{ij} + c_{ij})$ for all $i, j$

3. Identity element: $A + O = A$ where $O$ is the zero matrix (all entries 0)

   • Why? $a_{ij} + 0 = a_{ij}$

4. Additive inverse: $A + (-A) = O$ where $-A = [-a_{ij}]$

   • Why? $a_{ij} + (-a_{ij}) = 0$

Subtraction as addition: $A - B = A + (-B)$, so subtraction inherits these properties (except comutativity becomes anti-comutativity: $A - B =-(B - A)$).

[!example] Worked Example 1: Valid Addition

Given: $A = \begin{bmatrix} 2 & -3 & 5 \\ 1 & 0 & -2 \end{bmatrix}, \quad B = \begin{bmatrix} 4 & 1 & -1 \\ -3 & 7 & 6 \end{bmatrix}$

Find: $A + B$

Solution:

Step 1: Check dimensions

• $A$ is $2 \times 3$
• $B$ is $2 \times 3$
• ✓ Identical, operation defined

Step 2: Add corresponding entries $A + B = \begin{bmatrix} 2+4 & -3+1 & 5+(-1) \ 1+(-3) & 0+7 & -2+6 \end{bmatrix}$

Why this step? Each entry $(i,j)$ in the result comes from $(i,j)$ in $A$ plus $(i,j)$ in $B$.

Step 3: Simplify $A + B = \begin{bmatrix} 6 & -2 & 4 \\ -2 & 7 & 4 \end{bmatrix}$

[!example] Worked Example 2: Valid Subtraction

Given: $P = \begin{bmatrix} 5 & -1 \ 3 & 2 \\0 & 4 \end{bmatrix}, \quad Q = \begin{bmatrix} 2 & 3 \\ -1 & 1 \\ 5 & -2 \end{bmatrix}$

Find: $P - Q$

Solution:

Step 1: Check dimensions

• Both are $3 \times 2$
• ✓ Valid

Step 2: Subtract element-wise $P - Q = \begin{bmatrix} 5-2 & -1-3 \ 3-(-1) & 2-1 \\ 0-5 & 4-(-2) \end{bmatrix}$

Why this step? Subtraction means adding the additive inverse: $p_{ij} + (-q_{ij})$.

Step 3: Result $P - Q = \begin{bmatrix} 3 & -4 \\ 4 & 1 \\ -5 & 6 \end{bmatrix}$

[!example] Worked Example 3: Invalid Operation

Given: $M = \begin{bmatrix} 1 & 2 & 3 \\ 4 & 5 & 6 \end{bmatrix}, \quad N = \begin{bmatrix} 7 & 8 \ 9 & 10 \\ 11 & 12 \end{bmatrix}$

Question: Can we compute $M + N$?

Solution:

Step 1: Check dimensions

• $M$ is $2 \times 3$ (2 rows, 3 columns)
• $N$ is $3 \times 2$ (3 rows, 2 columns)
• ✗ Not identical

Step 2: Conclude
$M + N$ is undefined.

Why? The $(1, 3)$ entry of $M$ (which is 3) has no corresponding entry in $N$ (which only has 2 columns). The $(3, 1)$ entry of $N$ (which is 11) has no partner in $M$ (which only has 2 rows).

[!mistake] Common Mistakes

Mistake 1: "Same number of total elements is enough"

Wrong thinking: Matrix $A$ is $2 \times 3$ (6 elements), Matrix $B$ is $3 \times 2$ (6 elements), so I can add them.

Why it feels right: Both have 6 numbers, seems like you could pair them up.

The fix: Position matters, not just count. A $2 \times 3$ matrix represents2 row-categories × 3 column-categories. A $3 \times 2$ represents 3 × 2 — different structure. Entry $(1, 3)$ in $A$ means "row-1, column-3" which doesn't exist in the $3 \times 2$ structure of $B$.

Steel-man: The confusion arises because we can reshape matrices in some contexts (vectorization). But addition requires structural alignment — you're not just combining numbers, you're combining relationships.

Mistake 2: "I can transpose one matrix to match dimensions"

Wrong thinking: If $A$ is $2 \times 3$ and $B$ is $3 \times 2$, compute $A + B^T$ and call it matrix addition.

Why it feels right: $B^T$ flips $B$ to $2 \times 3$, now dimensions match!

The fix: $A + B^T$ is a different operation than $A + B$. Transposing changes which relationships you're encoding. If $B$ tracks "products × stores", then $B^T$ tracks "stores × products" — the meaning of each entry has changed. You can compute $A + B^T$ if it makes sense for your problem, but it's not "matrix addition of $A$ and $B$" — it's "matrix addition of $A$ and the transpose of $B$".

Steel-man: In applied work (machine learning, physics), we do strategically transpose to enable operations. But we do so intentionally, aware we're changing the semantic structure. Don't transpose just to force arithmetic.

[!recall]- Feynman Explanation (To a 12-Year-Old)

Imagine you and your friend both have a chore chart — a grid with days across the top and chores down the side. Each box tells you how many minutes you spent on that chore that day.

Adding matrices is like combining your charts: "On Monday for dishes, I spent 15 minutes, you spent 20, so together we spent 35." You go box-by-box, adding the numbers.

But here's the catch: You can only do this if your charts have the exact same layout. If your chart has 7 days but your friend's has only 5 days, what do you put for days6 and 7? There's no number to add from their chart — it's blank, undefined!

That's why matrices need identical dimensions — every box in one chart needs a partner box in the other chart. If the grids don't match, you're stuck.

Subtraction is the same idea: "I spent 15 minutes, you spent 20, so you did 5 more minutes than me." Still need matching grids.

[!mnemonic] Memory Aid

"Same Size to Synthesize"

• Same Size: Dimensions must match exactly ($m \times n$ for both)
• Synthesize: Combine element-by-element to create the result

Visual: Picture two identical photo frames overlaying — each pixel adds to the pixel below it. Mismatched frames leave gaps.

Connections

• Matrix Notation and Terminology — prerequisite: understanding order/dimensions
• Zero Matrix and Identity Matrix — the additive identity $O$ comes from here
• Scalar Multiplication of Matrices — can be combined: $k(A + B) = kA + kB$
• Matrix Multiplication — contrast: multiplication has different dimension requirements ($m \times n$ times $n \times p$, not $m \times n$ for both)
• Transpose of a Matrix — how $A + B^T$ differs from $A + B$
• System of Linear Equations — matrix addition models combining systems (e.g., economic scenarios)

Flashcards

What is the necessary condition for two matrices to be added or subtracted? :: The matrices must have identical dimensions (same number of rows AND same number of columns, i.e., same order $m \times n$).

If $A$ is a $3 \times 4$ matrix and $B$ is a $3 \times 5$ matrix, is $A + B$ defined? :: No. Although both have 3 rows, $A$ has 4 columns while $B$ has 5 columns. Dimensions are not identical.

Concept Map

Hinglish (regional understanding)