Visual walkthrough — Solving systems using matrix inversion

2.6.12 · D2 · Maths › Matrices & Determinants — Introduction › Solving systems using matrix inversion

Yeh parent recipe hai jo dheere-dheere, ek picture per step, draw ki gayi hai. Hum shuru karte hain do sadharan equations se aur end karte hain ek single boxed formula par — lekin har symbol apni jagah earn karta hai pehle use hone se, aur har move ko ek picture milti hai.

Hum poore time ek hi running example use karenge: $2x+3y=5,x-y=1.$

Ek flat sheet par do seedhi lines. Jahan woh cross karti hain wahi answer hai. Dekho kaise matrices us crossing point ko ek clean formula mein repackage karti hain.

---

Step 1 — Do equations matlab do lines

KYA kiya hamne: har equation ko ek line mein badla. KYU: ek solution ko dono promises satisfy karni chahiye, isliye woh dono lines par hona chahiye. Geometry algebra ki jagah le leti hai — ab hum ek intersection dhundh rahe hain. KAISA DIKHTA HAI: do chalk lines ek dot par cross karti hain.

Figure mein woh dot wahi answer hai jise hum dhundh rahe hain: $(x,y)=(8/5,3/5)$. Neeche sab kuch woh machinery hai jo us dot ko graph eyeball kiye bina compute karti hai.

---

Step 2 — Numbers ko ek grid $A$ mein, ek stack $X$ mein, ek stack $B$ mein dalna

Koi bhi symbol aane se pehle, yeh raha har grid kya hai.

$A$ ko term by term padhna: upar wali row $\left[\begin{array}{cc}2& 3\end{array}\right]$ pehli equation ke coefficients hain ($2x+3y$); neeche wali row $\left[\begin{array}{cc}1& -1\end{array}\right]$ doosri ke hain ($1x-1y$). Column $B$ mein "$=5$" aur "$=1$" hain.

KYA kiya hamne: numbers ko letters se alag kar diya. KYU: akele numbers (grid $A$) system ki poori structure carry karte hain. Baad mein hum grid par kaam kar sakte hain bina $x,y$ ko saath ghisaate. KAISA DIKHTA HAI: har equation ko colour karo aur dekho uske numbers ek row mein march karte hain.

---

Step 3 — "Row times column" ka matlab kya hai (taaki $AX$ system ko rebuild kare)

Humne abhi tak ek grid ko ek column se multiply nahi kiya, toh yahan ek hi rule hai, ek picture se bana.

$A$ ki dono rows ke liye yeh karo toh do results ek column mein stack ho jaate hain: $AX=\left[\begin{array}{cc}2& 3\\ 1& -1\end{array}\right]\left[\begin{array}{c}x\\ y\end{array}\right]=\left[\begin{array}{c}2x+3y\\ x-y\end{array}\right].$

Ab us column ko $B$ ke barabar set karo: $\left[\begin{array}{c}2x+3y\\ x-y\end{array}\right]=\left[\begin{array}{c}5\\ 1\end{array}\right]⟺\{\begin{array}{l}2x+3y=5\\ x-y=1\end{array}.$

KYA kiya hamne: dikhaya ki $AX=B$ literally wahi do original equations hain. KYU: yahi poori repackaging ka justification hai — compact form mein kuch khota nahi. Row $i$ times $X$ exactly equation $i$ reproduce karta hai. KAISA DIKHTA HAI: ek row column ke neeche slide karti hai, terms pair karti hai, phir answer column ke ek slot mein gir jaati hai.

---

Step 4 — $A$ ek machine hai jo points ko move karti hai

Toh poora sawaal yeh ban jaata hai: "woh kaun sa arrow $X$ hai jise machine $A$ arrow $B$ par bhejti hai?"

KYA kiya hamne: "system solve karo" ko "ek machine ka input reverse-engineer karo" ke roop mein reframe kiya. KYU: kyunki jis machines ko reverse kiya ja sake unke paas ek clean undo button hota hai — aur woh button humein directly $X$ de dega. KAISA DIKHTA HAI: left par ek input arrow, machine $A$, right par ek output arrow. Hum output $B$ jaante hain; input missing hai.

---

Step 5 — Undo machine $A^{-1}$

$A^{-1}A=I$ ka matlab "undo" kyun hai? Kyunki pehle $A$ lagana phir $A^{-1}$ lagana bilkul wahi hai jaise do-nothing machine lagana — tum wahan pahunch jaate ho jahan se shuru kiya tha.

KYA kiya hamne: reverse machine aur do-nothing machine ko naam diya. KYU: hum "matrix se divide" nahi kar sakte, lekin hum ek machine ko undo kar sakte hain. $A^{-1}$ matrix ki duniya mein dividing ka version hai. KAISA DIKHTA HAI: output arrow $B$ ko $A^{-1}$ ke through ulta daala ja raha hai, input arrow $X$ par land karta hua.

Undo machine ki concrete recipe (dekho Adjoint and Inverse of a Matrix): $A^{-1}=\frac{1}{\det A}\mathrm{adj}(A).$ Yahan $\det A$ ek single number hai jo "machine kitna area stretch ya squash karti hai" measure karta hai (dekho Determinants), aur $\mathrm{adj}(A)$ cofactor matrix ka transpose hai (dekho Cofactors and Minors).

---

Step 6 — Derivation, ek line at a time

Ab algebra, har move justified.

Packed form se shuru karo: $AX=B.$ Yahan $A$, $X$ ke left par baitha hai. Yaad raho ki matrices ke liye order matter karta hai — $A^{-1}A$ cancel hota hai lekin $A{A}^{-1}$ ek alag order hai, isliye hum dhyan rakhenge ki hum kis side act karte hain.

Dono sides ko left par $A^{-1}$ se multiply karo: $A^{-1}(AX)=A^{-1}B.$ Humne left choose kiya kyunki $A^{-1}$ ko seedha $A$ ke paas baithna chahiye jise woh cancel kar raha hai. Right se daalenge toh $A^{-1}$ ke paas cancel karne ke liye kuch nahi hoga.

Regroup karo (matrix multiplication associative hai — brackets shift ho sakte hain): $(A^{-1}A)X=A^{-1}B.$

$A^{-1}A$ ko $I$ se replace karo (Step 5 ki definition), aur $IX$ ko $X$ se (do-nothing machine): $X=A^{-1}B.$

KYA kiya hamne: $A$ ko $X$ se undo kiya, $A^{-1}$ se cancel karke. KYU har step: left-multiply → $A^{-1}$ ko $A$ ke samne laane ke liye; regroup → $A^{-1}A$ ko saath laane ke liye; simplify → kyunki $A^{-1}A=I$ aur $I$ kuch nahi karta. KAISA DIKHTA HAI: $A$ aur $A^{-1}$ ek doosre ko annihilate karte hain, left par akela $X$ reh jaata hai.

Hamare numbers daalo ($\det A=2(-1)-3(1)=-5$):

X=\frac{1}{-5}\begin{bmatrix}-1&-3\\-1&2\end{bmatrix}\begin{bmatrix}5\\1\end{bmatrix}=\frac{1}{-5}\begin{bmatrix}-8\\-3\end{bmatrix}=\begin{bmatrix}8/5\\3/5\end{bmatrix}.$$ Yahi exactly woh crossing dot hai Step 1 se. ✓ --- ## Step 7 — Degenerate case: jab machine space ko flatten kar de Upar sab kuch $\det A\neq0$ pe dependent tha. Dekho kya toot-ta hai warna. > [!intuition] > $\det A$ measure karta hai ki machine area ko kitna badalta hai. Agar $\det A=0$, toh machine **poore plane ko ek single line par squash kar deti hai** — har point us line par collapse ho jaata hai. Jab space flat ho jaaye, do alag inputs ek hi output par land kar sakte hain, isliye koi unique undo nahi hoti — $A^{-1}$ exist nahi karta (tum $\det A=0$ se divide kar rahe hote). Do sub-cases system ki fate decide karte hain. $AX=B$ ko $\operatorname{adj}(A)$ se multiply karo aur identity $\operatorname{adj}(A)\,A=(\det A)I$ use karo: $$(\det A)\,X=(\operatorname{adj}A)\,B.$$ Jab $\det A=0$ toh left side zero column $O$ hai, forcing $(\operatorname{adj}A)B=O$: > [!definition] $\det A=0$ fork > - $(\operatorname{adj}A)B\neq O$ → **impossible** → ==no solution== (lines parallel hain, kabhi nahi milti). > - $(\operatorname{adj}A)B= O$ → ==infinitely many solutions== (dono lines ek hi line hain) — ya koi nahi; substitution se confirm karo. **KYA** kiya hamne: woh case cover kiya jise boxed formula handle nahi kar sakta. **KYU**: reader ko kabhi ek unsolvable input pe surprise nahi hona chahiye — ab har scenario dikhaya ja chuka hai. **KAISA DIKHTA HAI**: left panel, do parallel lines jo kabhi cross nahi karti (no solution); right panel, do lines jo exactly ek doosri ke upar bichhi hain (infinitely many). Dekho [[Consistency of Linear Systems]]. > [!mistake] $\det A$ check kiye bina $A^{-1}$ ke liye haath badhana > **Kyun tempt karta hai:** recipe universal lagti hai. > **Fix:** $A^{-1}$ $\det A$ se divide karta hai. Hamesha **pehle** $\det A$ compute karo. Agar zero hai, inversion chhoddo aur upar wala fork use karo. --- ## Ek-picture summary Poora safar ek board par: system → grid $AX=B$ → machine view → $A$ ko $A^{-1}$ se cancel karo → $X=A^{-1}B$, saath mein $\det A=0$ trapdoor side mein drawn. > [!recall]- Feynman: poora walkthrough kisi dost ko batao > Hamare paas do straight lines thi aur hum unka crossing point chahte the. Pehle humne notice kiya ki har equation ek line hai, aur answer wahan milta hai jahan woh milti hain. Phir humne numbers ko ek tidy grid $A$ mein strip kiya, unknowns ko ek column $X$ mein stack kiya, aur answers ko ek column $B$ mein — aur check kiya ki "grid times column", pairing-and-adding karke, original equations ko exactly rebuild karta hai, isliye kuch kho nahi gaya. Aage humne $A$ ko ek machine picture kiya jo ek input arrow ko ek output arrow par push karti hai; hum output $B$ jaante hain lekin input $X$ nahi. Agar machine ko reverse kiya ja sake, toh uska undo button $A^{-1}$ hai, jo is tarah define hai ki $A$ karna phir $A^{-1}$ karna tumhe wahan chhodta hai jahan se shuru kiya. $X$ akele paane ke liye humne $AX=B$ ko left se $A^{-1}$ se multiply kiya — *left*, kyunki $A^{-1}$ ko us $A$ ko touch karna hai jise woh cancel karta hai — aur nikal aaya $X=A^{-1}B$. Aakhir mein, warning: agar machine plane ko flatten kar de ($\det A=0$) toh tum ise undo nahi kar sakte. Tab ya toh lines parallel hain (koi answer nahi) ya bilkul ek hi line hain (bemisal answers). > [!mnemonic] > **Grid it, machine it, undo it: $X=A^{-1}B$ — lekin tabhi jab $\det A$ zero na ho.** --- ## Connections - [[Solving systems using matrix inversion]] — woh parent recipe jise yeh page draw karta hai. - [[Determinants]] — woh $\det A$ jo decide karta hai ki machine reversible hai ya nahi. - [[Adjoint and Inverse of a Matrix]] — undo machine $A^{-1}$ actually kaise banti hai. - [[Cofactors and Minors]] — $\operatorname{adj}(A)$ ke andar ke ingredients. - [[Consistency of Linear Systems]] — Step 7 mein $\det A=0$ fork. - [[Cramer's Rule]] — ek sibling solver jo usi $\det A\neq0$ gate se guzarta hai. - [[Gaussian Elimination]] — bade grids ke liye practical alternative.