Linear independence — formal definition, testing: Visual walkthrough

4.5.15 · D2 · Maths › Linear Algebra (Full) › Linear independence — formal definition, testing

Step 1 — Vector kya hai, aur ek arrow kya kar raha hai?

KYA. $R^{2}$ mein ek vector bas ek number pair hota hai, jaise $(3, 1)$ . Hum isse ek arrow ke roop mein dekhte hain jo origin (point $(0, 0)$ ) se shuru hota hai aur us coordinate pair par khatam hota hai. Pehla number batata hai kitna daayein, doosra kitna upar.

KYUN. Jo kuch aage aata hai — "vectors ko milana", "ek point tak pahunchna" — yeh sab tab zyada samajh mein aata hai jab aap isse aise arrows ke roop mein dekhein jo aap ek ke baad ek rakh sakte ho. Koi bhi algebra aane se pehle yeh picture dimag mein pakki honi chahiye.

PICTURE. Laal arrow hai $v_{1} = (3, 1)$ : 3 daayein, 1 upar. Cyan arrow hai $v_{2} = (1, 2)$ .

Figure — Linear independence — formal definition, testing

Step 2 — Linear combination kya hoti hai? (arrows ko scale aur add karna)

KYA. Ek number $c$ lo (ek scalar) aur ek vector $v$ lo. Tab $c v$ arrow ko $c$ factor se stretch karta hai: $2 v$ do guna lamba hai, $- 1 v$ isse ulta kar deta hai, $0 v$ isse $0$ mein crush kar deta hai. Do vectors ko add karne ka matlab hai unhe tip to tail rakhna.

$c_{1} v_{1} + c_{2} v_{2}$

Term by term: $c_{1}$ , $v_{1}$ ko stretch karta hai; $c_{2}$ , $v_{2}$ ko stretch karta hai; aur $+$ doosre stretched arrow ko pehle ki tip par rakhta hai. Jahan aakhiri tip hoti hai, wahan aap pahunchte ho.

KYUN. Independence ki definition poori tarah combinations ke $0$ ke barabar hone ke baare mein hai. Isliye pehle physically dekhna zaroori hai ki combination kaisi lagti hai — stretches ki ek recipe jo aapko kisi destination tak le jaati hai.

PICTURE. Amber dashed arrows $1.5 v_{1}$ aur $- 0.5 v_{2}$ ko tip to tail rakhte hain; black arrow woh hai jahan unka sum land karta hai.

Step 3 — Khaas destination: origin par wapas aana

KYA. Ab hum woh key sawaal poochte hain: kya koi combination aapko wapas $0$ par pahuncha sakti hai? Matlab, kya aap scalars choose kar sakte ho taaki $c_{1} v_{1} + c_{2} v_{2} = 0 ?$ Iska hamesha ek boring tarika hota hai: $c_{1} = 0, c_{2} = 0$ — koi kadam nahi, aap origin se kabhi nikle hi nahi. Yahi trivial solution hai.

KYUN. Nonzero steps ke saath $0$ par wapas aana matlab hai arrows secretly cancel ho rahe hain — ek arrow ki stretch doosron se undo ho jaati hai. Yahi cancellation exactly redundancy hai. Isliye " $0$ tak exciting tarike se pahuncha ja sakta hai?" yahi independence ka sawaal hai.

PICTURE. Left: independent arrows — $0$ par wapas jaane ka ek hi tarika hai: koi kadam nahi. Right: do parallel arrows — aap ek ke saath bahar ja sakte ho aur doosre ke saath wapas aa sakte ho, $0$ tak ek nonzero round trip.

Step 4 — Nonzero round trip = "ek vector baaki ka copy hai"

KYA. Maano ek exciting solution exist karta hai, jaise $c_{1} \neq = 0$ . Poori equation ko $c_{1}$ se divide karo aur terms ko aaage-peeche karo: $c_{1} v_{1} + c_{2} v_{2} = 0 ⟹ v_{1} = - \frac{c _{2}}{c _{1}} v_{2} .$ Isko padhein: $v_{1}$ barabar hai $v_{2}$ ki ek scaled copy ke. Number $- \frac{c _{2}}{c _{1}}$ sirf stretch factor hai. Toh $v_{1}$ koi naya direction nahi laya tha — woh ek passenger tha.

KYUN. Yeh algebra (equation $= 0$ ) aur intuition ("koi vector doosron se nahi banta") ke beech ka bridge hai jo parent note ne promise kiya tha. Isse geometrically dekhne se yeh undeniable ho jaata hai.

PICTURE. Laal arrow $v_{1}$ bilkul cyan arrow $v_{2}$ ki line par baitha hai — same line, alag length. Redundancy visible ho gayi.

Step 5 — Sawaal ko matrix mein pack karna

KYA. Vectors ko matrix $A$ ke columns ke roop mein stack karo aur scalars ko column $c = (c_{1}, c_{2})$ mein collect karo. Tab matrix–vector product hi linear combination hai: $A c = ∣ v_{1} ∣ ∣ v_{2} ∣ [c_{1} c_{2}] = c_{1} v_{1} + c_{2} v_{2} .$ Toh independence ka sawaal ban jaata hai: kya $A c = 0$ ka sirf $c = 0$ solution hai? — ek homogeneous system.

KYUN. Arrows draw karna 2 ya 3 vectors ke liye kaam karta hai, lekin higher dimensions mein machinery chahiye. Ek matrix computer ko (ya row reduction ko) same geometric sawaal answer karne deta hai. Kuch naya nahi ho raha — sirf Steps 2–4 ko repackage kiya ja raha hai.

PICTURE. Donon arrows andar slide hokar $A$ ke donon columns ban jaate hain; label dikhata hai ki column $\times$ scalar combination ko kaise rebuild karta hai.

Step 6 — Pivots ginana: asli test

KYA. $A$ ko row-reduce karke staircase form mein laao aur pivots gino (har nonzero row ki leading nonzero entry). Jis column mein koi pivot nahi hai woh ek free variable correspond karta hai — ek knob jise aap freely ghuma sakte ho, jo infinitely many nonzero solutions deta hai.

Har column mein pivot $\Rightarrow$ $rank (A) = n$ $\Rightarrow$ sirf $c = 0$ $\Rightarrow$ independent.
Pivotless column $\Rightarrow$ free variable $\Rightarrow$ exciting solutions exist karte hain $\Rightarrow$ dependent.

KYUN. Yeh " $0$ tak exciting tarike se pahuncha ja sakta hai?" ko ek finite, mechanical count mein badal deta hai. Dekho rank — yeh hi independent columns ki count hai.

PICTURE. Do staircases: left mein har column mein pivot hai (independent); right mein ek zero row hai, isliye uska aakhiri column pivotless hai — ek free variable, dependence.

Step 7 — Square shortcut: determinant = signed area

KYA. Jab aapke paas exactly $n$ vectors hon $R^{n}$ mein (ek square $A$ ), toh determinant un arrows ka (signed) area/volume measure karta hai jo woh enclose karte hain. Agar woh dependent hain toh woh ek line/plane par aate hain — enclosed area $0$ ho jaata hai. $det A \neq = 0 ⟺ independent ⟺ A hai [[I n v er t ib l e ma t r i x t h eor e m ∣ in v er t ib l e]] .$

KYUN. Yeh ek single-number test hai sirf tab jab square ho. $2 \times 2$ ke liye term by term: $det [a c b d] = a d - b c$ woh parallelogram ka area hai jo donon columns span karte hain; area $0$ matlab woh parallel hain (Step 4 ki picture).

PICTURE. Left: independent pair ek asli parallelogram span karta hai (amber fill, area $\neq = 0$ ). Right: parallel pair — parallelogram bilkul flat ho gaya, area $0$ .

Step 8 — Har edge case, draw karke

KYA. Teen scenarios jinhein pivot count automatically handle karta hai:

Zero vector present hai. Agar $v_{1} = 0$ hai, toh $1 \cdot 0 = 0$ ek exciting solution hai ( $c_{1} = 1 \neq = 0$ ). Hamesha dependent.
Bahut zyada vectors ( $n >$ dimension). $A$ short aur wide hai, isliye zyada se zyada (dimension) pivots hain zyada columns ke liye $\Rightarrow$ ek pivotless column hamesha exist karta hai. Hamesha dependent — aapke paas space se zyada independent directions nahi ho sakte.
Ek single nonzero vector. Ek arrow jo $0$ nahi hai, $0$ tak sirf $c = 0$ se pahunch sakta hai. Hamesha independent.

KYUN. Parent note ne yeh rules list kiye the; yahan aap dekhte hain kyun har ek forced hai, taaki aap kabhi kisi case se unprepared na milein.

PICTURE. Teen panels:납작 $0$ arrow; ek 2-row matrix mein teen fat columns jo ek guaranteed free column ke saath; ek akela nonzero arrow jo sirf trivially cancel hota hai.

Ek picture mein summary

Ek figure, poori chain: arrows → combination → $0$ par wapas? → $A$ ke columns → pivots → independent ya dependent, saath mein square-case determinant shortcut branching off karta hai.

Recall Feynman Retelling — Walkthrough simple shabdon mein

Page ke beech se arrows imagine karo. "Combination" ek recipe hai: ise itna stretch karo, use itna stretch karo, unhe tip to tail rakho, dekho kahan pahuncho. Poora game ek sawaal hai: kya koi recipe jo real (nonzero) stretches leti hai aapko exact centre par wapas le ja sakti hai? Agar ghar jaane ka ek hi tarika hai "koi kadam nahi lo", toh aapke arrows sach mein alag-alag directions mein khich rahe hain — independent. Lekin agar aap ek arrow ke saath bahar ja sakte ho aur doosron ke saath wapas aa sakte ho, toh unmen se ek disguise mein copy thi — dependent, ek passenger. Drawing ke bina check karne ke liye, arrows ko ek box $A$ ke columns ke roop mein khada karo aur row-reduce karo: staircase mein "steps" (pivots) gino. Har column ke liye ek pivot matlab koi free knob nahi, toh sirf do-nothing recipe zero tak pahunchti hai — independent. Ek missing pivot ek free knob hai — ise ghoomao aur zero tak ek asli round trip milega — dependent. Jab box square ho toh bonus milta hai: determinant woh area hai jo arrows enclose karte hain, aur ek squashed-flat area (zero) "dependent" ke barabar hai. Aur tricky freebies: ek zero arrow hamesha passenger hota hai; space se zyada arrows mein hamesha passenger hoga; ek single non-zero arrow mein kabhi nahi hota.

Connections

Linear independence — formal definition, testing — woh parent jise yeh page visualise karta hai.
Homogeneous systems and null space — $A c = 0$ ; nonzero solution = dependence.
Rank of a matrix — Step 6 mein gine gaye pivots.
Determinant — Step 7 mein square-case area test.
Invertible matrix theorem — independent columns ⇔ invertible.
Basis and dimension · Span and spanning sets — independence + spanning = ek basis.