Identity, diagonal, symmetric, and orthogonal matrices

1.1.6 · AI-ML › Linear Algebra Essentials

1. Identity matrix

YE KYUN IMPORTANT HAI: $I$ multiplicative identity hai: $I A = A I = A$ . Ye matrix algebra ka " $1$ " hai.

5. Forecast-then-Verify

Recall Compute karne se pehle forecast karo

Q: Kya do diagonal matrices ka product hamesha diagonal hota hai? Forecast karo, phir verify karo. A: Haan. $(D_{1} D_{2})_{ij} = \sum_{k} (D_{1})_{ik} (D_{2})_{k j}$ ; nonzero ke liye chahiye $i = k$ aur $k = j$ , to $i = j$ . Off-diagonal entries zero ho jaate hain ⇒ diagonal, entries $d_{i}^{(1)} d_{i}^{(2)}$ ke saath.

Q: Kya do symmetric matrices ka product symmetric hota hai? A: Generally NAHI! $(A B)^{⊤} = B^{⊤} A^{⊤} = B A \neq = A B$ jab tak $A, B$ commute na karein. Neeche steel-manned kiya gaya hai.

Common mistakes

Flashcards

Identity matrix ko define karne wala entrywise formula kya hai?

I_{ij} = δ_{ij}

(Kronecker delta: 1 agar

i = j

, warna 0).

A

ko diagonal

D

se left-multiply karne se kya hota hai?

A

ki row

i

d_{i}

se scale hoti hai (kyunki

(D A)_{ij} = d_{i} A_{ij}

diag (d_{1}, \dots, d_{n})

ka determinant?

\prod_{i} d_{i}

Symmetric matrix ki definition?

A = A^{⊤}

, yaani

A_{ij} = A_{j i}

X^{⊤} X

hamesha symmetric kyun hai?

(X^{⊤} X)^{⊤} = X^{⊤} X

transpose-of-product rule se.

Orthogonal matrix ke liye defining condition?

Q^{⊤} Q = I

, equivalently

Q^{- 1} = Q^{⊤}

Orthogonal matrix kaunsi geometric property preserve karta hai?

Lengths aur angles (ye ek rotation/reflection hai).

Orthogonal

Q

ke liye

det Q

ki possible values?

\pm 1

(+1 rotation, −1 reflection).

Kya do symmetric matrices ka product symmetric hota hai?

Sirf tab jab wo commute karein (

A B = B A

); generally nahi.

Kisi bhi square matrix ko do named parts mein split karo.

A = \frac{1}{2} (A + A^{⊤})

symmetric

+ \frac{1}{2} (A - A^{⊤})

skew-symmetric.

Kaunsa theorem symmetric matrices ke liye real eigenvalues + orthonormal eigenbasis guarantee karta hai?

The Spectral Theorem.

Recall Feynman: 12-saal ke bacche ko samjhao

Socho ek stretchy graph paper ki sheet hai jisme dots hain.

Identity matrix paper ko bilkul waisa hi rehne deta hai — ek "kuch nahi karna" wala button.
Diagonal matrix paper ko sirf left-right aur up-down stretch karta hai, har direction ko apni apni amount se, kabhi tilt nahi karta.
Symmetric matrix kuch special tilted directions pe stretch karta hai lekin kabhi "twist" nahi karta — jaise dough ko seedha dabao, ghuma ke nahi.
Orthogonal matrix poori paper ko spin ya flip karta hai lekin kabhi stretch nahi karta, to har dot doosre dot se utni hi door rehta hai. Isliye computers unhe pasand karte hain: kuch bhi exaggerate nahi hota.

Connections

Matrix Multiplication and Transpose — rules $(A B)^{⊤} = B^{⊤} A^{⊤}$ jo poore note mein use hue.
Eigenvalues and Eigenvectors — diagonal aur symmetric seedha eigen-decomposition mein jaate hain.
Spectral Theorem and PCA — symmetric covariance ⇒ orthogonal eigenvectors.
Determinants and Invertibility — $det Q = \pm 1$ , $det D = \prod d_{i}$ .
QR Decomposition — $A = QR$ factor karo jisme $Q$ orthogonal ho.
Vector Norms and Dot Products — orthogonal maps ki length preservation.

Identity, diagonal, symmetric, and orthogonal matrices

1. Identity matrix

2. Diagonal matrix

3. Symmetric matrix

4. Orthogonal matrix

5. Forecast-then-Verify

Common mistakes

Flashcards

Connections

Concept Map