Visual walkthrough — Matrices — review, operations, types

4.5.6 · D2 · Maths › Linear Algebra (Full) › Matrices — review, operations, types

Hum sirf ek cheez assume karte hain: ek matrix ek machine hai jo arrows ko move karti hai. Baaki sab hum khud build karte hain.

---

Step 1 — Ek arrow bas do numbers hote hain

KYA. Graph paper draw karo. Ek vector $\mathbf{x}$ ek arrow hai origin (beech wala cross) se ek point tak. Usse poori tarah do numbers se describe kiya ja sakta hai: kitna daayein ($x_1$) aur kitna upar ($x_2$).

KYUN. Kisi machine ke arrow ko move karne se pehle, hume arrow ko numbers se naam dene ka tarika chahiye. Woh naming hi poora reason hai ki coordinates exist karte hain. Arrow-as-numbers idea ke liye Dot Product and Vectors dekho.

PICTURE. Neeche wala red arrow point $(3,1)$ tak pahunchta hai: teen steps daayein, ek step upar.

$\mathbf{x}=\left[\begin{array}{c}{x}_1\\ x_2\end{array}\right]\underset{\text{steps right}}{\underbrace{x_1}}\underset{\text{steps up}}{\underbrace{x_2}}$

---

Step 2 — Ek matrix ek arrow ko move karti hai: "apply $B$" ka matlab kya hai

KYA. Ek $2\times 2$ matrix $B$ arrow $\mathbf{x}$ ko leta hai aur ek naya arrow $B\mathbf{x}$ banata hai. Naye arrow ke pehle number ka rule hai: $B$ ki row 1 ke saath chalo, har entry ko $\mathbf{x}$ ki matching entry se multiply karo, unhe add karo. Row 2 ke liye bhi aisa hi.

$B\mathbf{x}=\left[\begin{array}{cc}{b}_{11}& b_{12}\\ b_{21}& b_{22}\end{array}\right]\left[\begin{array}{c}{x}_1\\ x_2\end{array}\right]=\left[\begin{array}{c}\stackrel{\text{new right-amount}}{\overbrace{b_{11}{x}_1+b_{12}{x}_2}}\\ \underset{\text{new up-amount}}{\underbrace{b_{21}{x}_1+b_{22}{x}_2}}\end{array}\right]$

Top slot mein term by term: $b_{11}$ kehta hai "purana right kitna naya right banta hai", $b_{12}$ kehta hai "purana up kitna naye right mein leak hota hai". Har entry ek mixing amount hai.

KYUN row-dot-vector? Kyunki ek matrix linear honi chahiye: input double karo toh output double ho, aur inputs add ho. Input se banane wala ek hi formula jo yeh satisfiy karta hai woh hai $x_1,x_2$ ka weighted sum — exactly ek row aur arrow ka dot product. Koi bhi curved rule linearity tod dega. Yeh machine Linear Transformations mein hai.

PICTURE. Neeche, pale-yellow arrow hai $\mathbf{x}=(1,1)$; $B$ use shear karta hai (daayein jhukta hai) chalk-blue arrow $B\mathbf{x}$ mein.

---

Step 3 — Do machines ek line mein: woh sawaal jo multiplication ko force karta hai

KYA. Maano hum pehle $B$ apply karte hain, phir result par $A$ apply karte hain. Toh arrow ka safar hai $\mathbf{x}\stackrel{B}{\to }B\mathbf{x}\stackrel{A}{\to }A(B\mathbf{x}).$

Sawaal: kya ek single matrix hai — ise $AB$ kaho — jo dono jumps ek saath kare? Agar haan, toh uski entries choose karne ke liye free nahi hain; woh wही hain jo $(AB)\mathbf{x}=A(B\mathbf{x})$ ko har arrow $\mathbf{x}$ ke liye sach banayein.

KYUN yeh, entrywise nahi? Hum matching entries multiply karke ek product invent kar sakte the (woh "Hadamard" idea parent ke mistake box mein hai). Lekin woh product kisi useful sawaal ka jawab nahi deta — woh "do $B$ then $A$" describe nahi karta. Hum woh product chahte hain jo machines ko chain kare, kyunki chaining wahi hai jo transformations, aur Systems of Linear Equations ke systems, actually karte hain.

PICTURE. Do-hop ka safar: pink arrow $\to$ (after $B$) blue arrow $\to$ (after $A$) yellow arrow. Hum ek aisi machine dhoondh rahe hain jo pink se seedha yellow tak jump kare.

---

Step 4 — Grind karo: row·column formula apne aap saamne aata hai

KYA. Aao $A(B\mathbf{x})$ honestly compute karein aur dekho ki $AB$ ki definition khud nikli. $\mathbf{y}=B\mathbf{x}$ likho, toh $y_k={\sum }_j{b}_{kj}{x}_j$ (Step 2 ka rule, jahan $k$ row index hai, $j$ input slots par sum hota hai).

Ab $A$ apply karo. $A\mathbf{y}$ ka $i$-th number hai row $i$ of $A$ dotted with $\mathbf{y}$:

$[A\mathbf{y}{]}_i={\sum }_k{a}_{ik}{y}_k={\sum }_k{a}_{ik}(\underset{y_k}{\underbrace{\sum _j{b}_{kj}{x}_j}}).$

Do sums ka order swap karo (allowed hai — same numbers alag order mein add ho rahe hain):

$[A(B\mathbf{x}){]}_i={\sum }_j(\underset{\text{call this }(AB{)}_{ij}}{\underbrace{\sum _k{a}_{ik}{b}_{kj}}})x_j.$

Term by term: $a_{ik}$ hai "slot $k$ kitna output $i$ ko feed karta hai"; $b_{kj}$ hai "input $j$ kitna slot $k$ ko feed karta hai". Unka product middle index $k$ par summed karke har path count karta hai input $j$ se output $i$ tak. Woh total hai $(AB{)}_{ij}$.

KYUN inner index match karna zaroori hai. Letter $k$ $A$ ke columns aur $B$ ke rows par simultaneously run karta hai. Agar $A$ ke $p$ columns hain lekin $B$ ke alag number of rows hain, toh sum ${\sum }_k$ ke terms mismatch ho jaate hain — yeh meaningless hai. Isliye inner dimensions agree karni chahiye.

PICTURE. Entry $(AB{)}_{ij}$ = $A$ ki highlighted row $B$ ke highlighted column ke saath slide karti hai; har aligned pair multiply hota hai, pile sum hota hai.

---

Step 5 — Ek real wala karo, entry by entry

KYA. Lo $A=\left[\begin{array}{cc}1& 2\\ 0& 3\end{array}\right]$, $B=\left[\begin{array}{cc}4& 0\\ 1& 5\end{array}\right]$. $AB$ ek slot at a time banao.

$(AB{)}_{11}=\underset{\text{row1⋅col1}}{\underbrace{1\cdot 4}}+2\cdot 1=6,(AB{)}_{12}=1\cdot 0+2\cdot 5=10,$ $(AB{)}_{21}=0\cdot 4+3\cdot 1=3,(AB{)}_{22}=0\cdot 0+3\cdot 5=15.$

$AB=\left[\begin{array}{cc}6& 10\\ 3& 15\end{array}\right].$

KYUN inhe is tarah line up karo. Slot $(2,1)$ use karta hai row 2 of $A$ aur column 1 of $B$ — row index $A$ se copy hota hai, column index $B$ se copy hota hai. Yahi bookkeeping poori skill hai.

PICTURE. Slot $(2,1)$ lit up: row 2 of $A$ column 1 of $B$ ke saamne, jo deta hai $0\cdot 4+3\cdot 1=3$.

---

Step 6 — Order matters: $BA$ compute karo aur dekho kaise disagree karta hai

KYA. Same $A,B$, lekin ab $B$ doosre act karta hai:

$BA=\left[\begin{array}{cc}4& 0\\ 1& 5\end{array}\right]\left[\begin{array}{cc}1& 2\\ 0& 3\end{array}\right]=\left[\begin{array}{cc}4\cdot 1+0\cdot 0& 4\cdot 2+0\cdot 3\\ 1\cdot 1+5\cdot 0& 1\cdot 2+5\cdot 3\end{array}\right]=\left[\begin{array}{cc}4& 8\\ 1& 17\end{array}\right].$

Compare karo: $AB=\left[\begin{array}{cc}6& 10\\ 3& 15\end{array}\right]\ne \left[\begin{array}{cc}4& 8\\ 1& 17\end{array}\right]=BA$.

KYUN yeh differ karte hain. $AB$ matlab hai "do $B$, then $A$"; $BA$ matlab hai "do $A$, then $B$". Yeh alag-alag safar hain — pehle moje phir joote vs pehle joote phir moje. Algebra simply record karta hai ki geometry alag hai.

PICTURE. Same starting arrow ke do safar: ek path (shear-then-scale) kahin alag land karta hai doosre se (scale-then-shear).

---

Step 7 — Degenerate cases: zero, identity, aur vanishing products

KYA. Teen limiting machines jo tumhe kabhi surprise nahi karni chahiye.

• Identity $I=\left[\begin{array}{cc}1& 0\\ 0& 1\end{array}\right]$ har arrow ko untouched chodti hai: $I\mathbf{x}=\mathbf{x}$. Row 1 $x_1$ pick karta hai, row 2 $x_2$ pick karta hai — pure "kuch nahi karo".
• Zero matrix $0$ har arrow ko origin par crush karta hai: $0\mathbf{x}=0$.
• Zero divisors. Do nonzero machines compose ho kar zero machine ban sakti hain: $\left[\begin{array}{cc}1& 0\\ 0& 0\end{array}\right]\left[\begin{array}{cc}0& 0\\ 0& 1\end{array}\right]=\left[\begin{array}{cc}0& 0\\ 0& 0\end{array}\right].$ Right machine arrow ka sirf "up" part rakhti hai; left machine sirf "right" part rakhti hai. Inhe sequence mein karo aur kuch nahi bachta.

KYUN ordinary numbers ke saath aisa nahi ho sakta. Plain numbers mein $ab=0$ force karta hai $a=0$ ya $b=0$. Matrices yeh todte hain kyunki ek machine poori ek direction throw away kar sakti hai (yeh singular hai, $\det =0$; dekho Determinant and Inverse of a Matrix). Do machines jo complementary directions discard karti hain kuch nahi bachne deti.

PICTURE. Left machine saare arrows ko horizontal line par flatten karti hai; right machine vertical line par flatten karti hai. Pehle vertical squash, phir horizontal — arrow origin par pahunch jaata hai.

---

Ek-picture summary

Upar sab kuch ek single claim par collapse hota hai: row·column rule ek convention nahi hai, yeh "machines ko chain karne" ka unique fingerprint hai.

Recall Feynman retelling

Graph paper par arrows imagine karo. Ek matrix ek choti machine hai jo har arrow ko pakad ke uski tip move karti hai — lekin sirf "fair" tarike se: arrow double karo, move double ho; arrows add karo, moves add ho. Woh fairness force karti hai ki har output number input numbers ka weighted mix ho — ek row aur arrow ka dot.

Ab do machines ek line mein rakho: pehle $B$, phir $A$. Kya ek super-machine dono ke barabar hai? Haan — aur jab tum patiently ek arrow ko follow karte ho, us super-machine ke numbers nikle row of $A$ times column of $B$, middle mein summed. Kisi ne woh formula choose nahi kiya; woh "inhe sequence mein karo" se khud gira.

Kyunki "do $B$ then $A$" ek alag trip hai "do $A$ then $B$" se — pehle moje phir joote vs pehle joote phir moje — $AB$ aur $BA$ usually agree nahi karte. Aur kyunki ek machine poori ek direction ko nothing par flatten kar sakti hai, do nonzero machines multiply ho kar do-nothing-but-crush zero machine ban sakti hain. Identity woh aalsee machine hai jo kuch nahi choota.

Connections

• Linear Transformations — woh machines jinhe yeh matrices represent karti hain.
• Dot Product and Vectors — har product entry inhi mein se ek hai.
• Systems of Linear Equations — inhe chain aur invert karne se solve hote hain.
• Determinant and Inverse of a Matrix — kaun si machines ek direction flatten karti hain (singular).
• Eigenvalues and Eigenvectors — arrows jinhe ek machine sirf stretch karti hai.