Vector addition, scalar multiplication, and geometric interpretation

1.1.2 · AI-ML › Linear Algebra Essentials

1. Vector kya hota hai? (WHAT)

Column kyun? Kyunki matrix–vector multiplication (baad ke chapters mein) aise define ki gayi hai ki ek matrix left se column par act karta hai. Abhi consistency rakhne se baad mein takleef nahi hogi.

2. Vector Addition (HOW + WHY)

First principles se derivation

Ek vector ek displacement represent karta hai. Agar $u$ mujhe origin $O$ se point $A$ tak le jaata hai, aur $v$ ek aur displacement hai, toh unhe add karne ke liye main $A$ se $v$ ke saath aage badhta hun, aur $B$ par pahunchta hun.

Poore safar ka $x$ -displacement = ( $u$ ka x-move) + ( $v$ ka x-move) = $u_{1} + v_{1}$ .
Yahi logic har axis par lagti hai.

Isliye addition componentwise hi honi chahiye — yeh koi hat se nikali definition nahi hai, yeh "displacements har axis par independently accumulate hote hain" se forced hai.

Figure — Vector addition, scalar multiplication, and geometric interpretation

3. Scalar Multiplication (HOW + WHY)

Derivation

Scaling ka matlab hai "same displacement $c$ times karo." $v$ do baar karna = $v + v$ , jiske components $2 v_{1}, \dots$ hain. Integers se reals tak continuity se extend karne par $c v = (c v_{1}, \dots, c v_{n})$ milta hai.

4. Unhe combine karna: ML ka workhorse

5. Worked Examples

6. Common Mistakes (Steel-manned)

7. Feynman & Mnemonic

Recall Ek 12-saal ke bacche ko explain karo (reveal karne ke liye click karo)

Socho tum chal rahe ho. Ek vector ek instruction hai: "3 kadam east jao, 4 kadam north jao." Do instructions add karne ka matlab hai pehle wali walk karo, phir doosri — tum kahin pahunchte ho, aur woh final jagah sum hai. Scalar multiplication ek speed dial ki tarah hai: 2 se multiply karo = same direction mein do guni lambi walk; $- 1$ se multiply karo = same doori par ulta chalo. Bas itna hi hai — AI mein sab fancy cheezein sirf laakhon aisi walks ka addition hain.

8. Active Recall

Do vectors kaise add karte hain?

Componentwise:

(u + v)_{i} = u_{i} + v_{i}

; geometrically tip-to-tail.

Vector addition componentwise kyun hai (first principles)?

Har axis par displacements independently accumulate hote hain, isliye har coordinate alag add hota hai.

c v

kya hai?

Har component ko

c

se multiply karo:

(c v_{1}, \dots, c v_{n})

; arrow ko stretch/flip karta hai.

c v

ki length ka formula?

∥ c v ∥ = ∣ c ∣ ∥ v ∥

∥ c v ∥

mein

c

ki jagah

∣ c ∣

kyun?

Length non-negative hoti hai; negative scalar direction flip karta hai lekin length

∣ c ∣

se scale hoti hai.

Vectors ke liye triangle inequality batao.

∥ u + v ∥ \leq ∥ u ∥ + ∥ v ∥

, equality tab jab same direction ho.

Linear combination kya hota hai?

c_{1} v_{1} + \dots + c_{k} v_{k}

— vectors ka scaled sum; ek neuron ka core operation.

Do vectors add karne ke liye kya zaruri hai?

Unka dimension same hona chahiye.

\frac{1}{2} (u + v)

ka geometric meaning?

u

aur

v

ki tips ko join karne wale segment ka midpoint.

Kya vector addition commutative hai?

Haan:

u + v = v + u

(parallelogram = dono tip-to-tail orders ka same result).

9. Connections

Vectors and vector spaces — woh axioms jo ye operations follow karti hain
Linear combinations, span, and basis — seedha add + scale se banta hai
Dot product and vector norms — magnitude yahan se generalize hoti hai
Matrix-vector multiplication — matrix output uske columns ka linear combination hai
Neural network forward pass — weighted sum = scale + add, baar baar
Word embeddings — embeddings ko average/subtract karna inhi exact operations se hota hai