4.5.10 · Maths › Linear Algebra (Full)
Linear equations ka ek system bas ek machine hai jo variables ko mix karti hai . Solve karna matlab hai unmixing — ek ek variable ko alag karna. Row echelon form (REF) woh "staircase" hai jo hum banate hain taaki har nayi row ek naya fresh variable dikhaye. Reduced row echelon form (RREF) aur aage jaata hai: yeh staircase ke upar bhi clean karta hai, taaki answer seedha padha ja sake . REF = "back-substitute karne ke liye ready". RREF = "answer pehle se page par likha hai".
Definition Pivot / leading entry
Kisi bhi non-zero row mein, leading entry (pivot) woh pehla non-zero number hota hai jo left se right padhne par milta hai. Uski column ek pivot column kehlati hai.
Definition Reduced Row Echelon Form (RREF)
Ek matrix RREF mein hai agar woh REF mein hai aur :
4. Har pivot ==1 == ke barabar hai (ek leading 1 ).
5. Har pivot apni column mein single non-zero entry hai (upar aur neeche dono zeros hain).
Key fact: kisi bhi matrix ka RREF ==unique == hota hai. REF unique nahi hota (kaafi saari staircases ek hi answer tak pahunch sakti hain).
Intuition "Pivot ke neeche zero kyun hona chahiye"
Agar column j mein row i par ek pivot hai, aur hum uske neeche sab kuch 0 force kar dete hain, to koi bhi lower equation us variable ko ab contain nahi karti . Humne use top equations mein isolate kar diya hai. Yeh repeat karte karte, har variable upar ki taraf trap hota jaata hai → last pivot row seedha ek variable deta hai → upar ki taraf back-substitute karo. Staircase literally ek ordering hai ki pehle kise solve karein .
Sirf teen elementary row operations use karne ki ijazat hai, kyunki har ek solution set ko preserve karti hai :
Leftmost column dhoondo jisme non-zero entry ho → pivot column.
Kisi non-zero entry ko swap karke pivot position par lao (0 se divide karne se bachne ke liye).
Replace use karo taaki pivot ke neeche sab kuch zero ho jaaye.
Us row+column ko ignore karo, neeche-daayein submatrix par recurse karo.
Upar sab karo, phir bottom pivot se upar ki taraf jaate hue:
5. Har pivot row ko scale karo taaki pivot 1 ban jaaye.
6. Replace use karo taaki har pivot ke upar bhi entries zero ho jaayein.
Solve karo
⎩ ⎨ ⎧ x + 2 y + z = 6 2 x + 5 y + 3 z = 15 x + 3 y + 3 z = 11
Augmented matrix:
1 2 1 2 5 3 1 3 3 6 15 11
Step A: R 2 → R 2 − 2 R 1 , R 3 → R 3 − R 1 .
Yeh step kyun? Pivot top-left ka 1 hai; hum uske neeche wali entries ko khatam karte hain taaki x , equations 2 aur 3 se chala jaaye.
1 0 0 2 1 1 1 1 2 6 3 5
Step B: R 3 → R 3 − R 2 .
Yeh step kyun? Agla pivot row 2, column 2 mein 1 hai. Uske neeche wali entry khatam karo taaki y equation 3 se chala jaaye.
1 0 0 2 1 0 1 1 1 6 3 2 (REF — staircase done!)
REF se, back-substitute karo: z = 2 , phir y + z = 3 ⇒ y = 1 , phir x + 2 y + z = 6 ⇒ x = 2 .
Ab RREF ki taraf (pivots ke upar clean karo, bottom-up):
Step C: R 2 → R 2 − R 3 , R 1 → R 1 − R 3 .
Kyun? Row 3 ke pivot ki column clear karo.
1 0 0 2 1 0 0 0 1 4 1 2
Step D: R 1 → R 1 − 2 R 2 .
Kyun? Row 2 ke pivot ke upar clear karo.
1 0 0 0 1 0 0 0 1 2 1 2 (RREF)
Seedha padho: x = 2 , y = 1 , z = 2 . ✅
[ 1 2 2 4 3 7 4 9 ]
Step A: R 2 → R 2 − 2 R 1 .
Kyun? Leading 1 ke neeche khatam karo.
[ 1 0 2 0 3 1 4 1 ] ( REF )
Step B: R 1 → R 1 − 3 R 2 (column 3 mein pivot ke upar clean karo).
[ 1 0 2 0 0 1 1 1 ] ( RREF )
Pivot columns: 1 , 3 . Column 2 mein koi pivot nahi ⇒ y ek free variable hai.
x = 1 − 2 y , z = 1 , y free. Infinitely many solutions hain, ek poori line. Kyun? Variables se kam pivots hain ⇒ bachi hui freedom.
[ 1 2 1 2 2 5 ] R 2 → R 2 − 2 R 1 [ 1 0 1 0 2 1 ]
Row 2 padhi jaati hai 0 = 1 . Augmented (last) column mein ek pivot ⇒ no solution . Yeh kyun matter karta hai: constants column mein pivot algebraically yeh cheekh kar kehta hai "contradiction!".
Common mistake Common errors ko steel-man karke dekho
(1) "REF unique hoti hai." Kyun sahi lagta hai: RREF unique hai, toh REF bhi hogi. Fix: REF sirf staircase shape constrain karti hai; alag alag scalings/replacements alag alag REFs dete hain (e.g. pivots ka 1 hona zaroori nahi). Sirf RREF unique hoti hai.
(2) RREF ke liye pivots ke upar clear karna bhool jaana. Kyun sahi lagta hai: jab staircase neeche se clean lagti hai, tab lagta hai ho gaya. Fix: RREF mein pivots apni column mein sirf non-zero hone chahiye — upar bhi zeros hone chahiye.
(3) "Last column mein pivot theek hai." Kyun sahi lagta hai: woh bas ek aur pivot hai. Fix: ek augmented matrix mein last column constants hai; wahan pivot ka matlab hai 0 = nonzero ⇒ inconsistent.
(4) Zero row ko pivot row count karna. Fix: zero rows mein koi pivot nahi hota; rank = pivots ki sankhya = REF mein non-zero rows ki sankhya.
Recall Feynman: ek 12 saal ke bacche ko samjhao
Sochoo ek ganda blocks ka tower jisme har variable kaafi saare boxes mein chhupa hua hai. Hum boxes ko slide karte hain (row operations) taaki blocks ek staircase bana lein: top step mein saare variables hain, agla step mein ek kam, aur bottom step mein sirf ek raaz ka number hai. Jab ek baar tumhe bottom wala mil jaata hai, tum seedhi seedhaan upar chadho baaki fill in karte hue. RREF tab hota hai jab hum steps ko bhi dust off kar dete hain taaki har step ek bilkul sahi answer dikhaye — chadhne ki zaroorat nahi!
"Pivots March Right, Stand Alone as Ones" →
March Right = har pivot aur daayein (REF staircase),
Stand Alone = apni column mein single non-zero (RREF),
Ones = leading 1's (RREF).
Yeh bhi: R EF = R eady (back-substitute karne ke liye), R REF = R ead (seedha padho).
What is a pivot (leading entry)? Kisi non-zero row mein pehli non-zero entry.
Three conditions for REF? Zero rows sabse neeche; har pivot strictly upar wale se daayein; har pivot ke neeche zeros.
Two extra conditions for RREF beyond REF? Har pivot 1 hai; har pivot apni column mein single non-zero entry hai (upar bhi zeros hain).
Is REF unique? Nahi — sirf RREF kisi given matrix ke liye unique hoti hai.
What are the three elementary row operations? Rows swap karo; row ko non-zero constant se scale karo; ek row ka multiple doosri row mein add karo.
Why do row operations preserve solutions? Har ek reversible hai, isliye koi solution gain ya lost nahi hoti.
What does a pivot in the augmented (last) column mean? System inconsistent hai (0 = nonzero) — koi solution nahi.
What is a free variable? Woh variable jiske column mein koi pivot nahi hota; woh koi bhi value le sakta hai.
How is rank read from REF? Pivots ki sankhya = non-zero rows ki sankhya.
REF → solving method vs RREF → solving method? REF: back-substitution; RREF: solution seedha padho.
Gaussian vs Gauss–Jordan elimination? Gaussian REF tak pahunchata hai (pivots ke neeche zeros); Gauss–Jordan RREF tak continue karta hai (leading 1's, upar bhi zeros).
first non-zero entry defines
zeros below pivots produces
clean above and scale to 1
extends Gaussian to produce
Elementary row operations