4.5.33
Intuition Badi picture (KYU)Ordinary dot product u ⋅ v = u 1 v 1 + u 2 v 2 + ⋯ + u n v n teen kaam karta hai: length measure karta hai (∥ v ∥ = v ⋅ v angle measure karta hai (cos θ = ∥ u ∥∥ v ∥ u ⋅ v perpendicularity detect karta hai (u ⋅ v = 0 geometry ek single algebraic operation se aati hai. Toh agar hum weird spaces — polynomials, functions, random variables — par geometry chahte hain, toh bas ek aisa operation invent karna hoga jo dot product ki tarah behave kare. Woh operation jo rules follow kare, woh hain inner product ke axioms . Jo bhi unhe satisfy kare, usse automatically length, angle, aur orthogonality "free mein" mil jaati hai.
Maano V inner product ek function ⟨ ⋅ , ⋅ ⟩ : V × V → R u , v , w ∈ V c
Symmetry: ⟨ u , v ⟩ = ⟨ v , u ⟩ Linearity (1st slot mein): ⟨ c u + w , v ⟩ = c ⟨ u , v ⟩ + ⟨ w , v ⟩ Positive-definiteness: ⟨ v , v ⟩ ≥ 0 == ⟨ v , v ⟩ = 0 ⟺ v = 0 ==
Ek vector space jo aise function se equipped ho, woh ek inner product space
YEH TEEN AXIOMS KYU? Har ek geometry ka ek piece bachata hai:
Symmetry → u , v v , u
Linearity → "vector ko stretch karna uske projections scale karta hai", isse compute karna possible hota hai.
Positive-definiteness → guarantee karta hai ki ∥ v ∥ = ⟨ v , v ⟩ real, non-negative number hai, aur sirf zero vector ki length zero hoti hai.
Intuition Cauchy–Schwarz ki zaroorat kyun haicos θ = ∥ u ∥∥ v ∥ ⟨ u , v ⟩ [ − 1 , 1 ] exactly Cauchy–Schwarz inequality hai. Toh yeh inequality "admission ka price" hai taaki "angle" ka koi matlab ho.
Worked example Derivation: Cauchy–Schwarz∣ ⟨ u , v ⟩ ∣ ≤ ∥ u ∥ ∥ v ∥ Trick: vector u − t v har real t 0 ≤ ⟨ u − t v , u − t v ⟩ Yeh step kyun? Positive-definiteness (axiom 3) guarantee karta hai ki right side ≥ 0 0 ≤ ⟨ u , u ⟩ − 2 t ⟨ u , v ⟩ + t 2 ⟨ v , v ⟩ Kyun? Yeh t f ( t ) = ∥ v ∥ 2 t 2 − 2 ⟨ u , v ⟩ t + ∥ u ∥ 2 ≥ 0 t a t 2 + b t + c discriminant ≤ 0 hota hai:
b 2 − 4 a c ≤ 0 ⇒ 4 ⟨ u , v ⟩ 2 − 4∥ v ∥ 2 ∥ u ∥ 2 ≤ 0 Kyun? Agar discriminant positive hota toh parabola zero se neeche jaata, jo axiom 3 ko contradict karta.
⇒ ⟨ u , v ⟩ 2 ≤ ∥ u ∥ 2 ∥ v ∥ 2 ⇒ ∣ ⟨ u , v ⟩ ∣ ≤ ∥ u ∥∥ v ∥ ■
Worked example Derivation: Triangle inequality∥ u + v ∥ ≤ ∥ u ∥ + ∥ v ∥ ∥ u + v ∥ 2 = ⟨ u + v , u + v ⟩ = ∥ u ∥ 2 + 2 ⟨ u , v ⟩ + ∥ v ∥ 2 Yeh step kyun? Linearity se expand karo taaki cross term saamne aaye.
≤ ∥ u ∥ 2 + 2∥ u ∥∥ v ∥ + ∥ v ∥ 2 = ( ∥ u ∥ + ∥ v ∥ ) 2 Kyun? Humne ⟨ u , v ⟩ ≤ ∣ ⟨ u , v ⟩ ∣ ≤ ∥ u ∥∥ v ∥ ■
R 2 ⟨ u , v ⟩ = 3 u 1 v 1 + 2 u 2 v 2
Symmetry ✓ (multiplication commute karta hai). Linearity ✓ (distribute karta hai).
Positivity: ⟨ v , v ⟩ = 3 v 1 2 + 2 v 2 2 ≥ 0 v 1 = v 2 = 0 Weights matter kyun karte hain? Weights axes ko stretch karte hain — is space mein distances usual wali nahi hain. (Negative weights positivity tod denge → inner product nahi.)
Compute karo ∥ ( 1 , 1 ) ∥ = 3 ( 1 ) + 2 ( 1 ) = 5
Worked example 2. Evaluation ke zariye polynomials ka inner productDegree ≤ 1 ⟨ p , q ⟩ = p ( 0 ) q ( 0 ) + p ( 1 ) q ( 1 ) p ( x ) = x , q ( x ) = 1 − x ⟨ p , q ⟩ = p ( 0 ) q ( 0 ) + p ( 1 ) q ( 1 ) = ( 0 ) ( 1 ) + ( 1 ) ( 0 ) = 0 Yeh nice kyun hai? x 1 − x orthogonal hain! Polynomials par geometry.
Worked example 3. Function inner product (integral)[ 0 , 1 ] ⟨ f , g ⟩ = ∫ 0 1 f ( x ) g ( x ) d x sin ( 2 π x ) cos ( 2 π x ) ∫ 0 1 sin ( 2 π x ) cos ( 2 π x ) d x = 2 1 ∫ 0 1 sin ( 4 π x ) d x = 2 1 [ − 4 π c o s 4 π x ] 0 1 = 0 Yeh step kyun? sin A cos A = 2 1 sin 2 A Fourier series ka engine hai.
Common mistake "Koi bhi bilinear symmetric form inner product hoti hai."Yeh sahi kyun lagta hai: symmetry + linearity "zyaadatar rules" ki tarah lagte hain, toh log positive-definiteness skip kar dete hain.
Fix: ⟨ u , v ⟩ = u 1 v 1 − u 2 v 2 ⟨( 0 , 1 ) , ( 0 , 1 )⟩ = − 1 < 0 negotiate nahi hoti .
cos θ 1.4 Yeh sahi kyun lagta hai: tum bas formula mein plug karte ho.
Fix: Cauchy–Schwarz ∣ cos θ ∣ > 1 forbid karta hai. Agar > 1
Common mistake "Orthogonal ka matlab hamesha usual right angle hota hai."Yeh sahi kyun lagta hai: R 2 Fix: x 1 − x 90° Orthogonality chosen inner product ke relative hoti hai.
Real inner product ke teen axioms Symmetry, linearity in first slot, positive-definiteness (⟨ v , v ⟩ ≥ 0 = 0 ⟺ v = 0
Inner product se norm kaise define hoti hai Cauchy–Schwarz ka statement ∣ ⟨ u , v ⟩ ∣ ≤ ∥ u ∥ ∥ v ∥
Cauchy–Schwarz derive karne ki key trick ⟨ u − t v , u − t v ⟩ ≥ 0 t ≤ 0
Discriminant ≤ 0 Ek quadratic in t b 2 − 4 a c ≤ 0
Inner product space mein angle ki definition cos θ = ∥ u ∥∥ v ∥ ⟨ u , v ⟩ [ − 1 , 1 ]
Do vectors orthogonal kab hote hain Jab ⟨ u , v ⟩ = 0
u 1 v 1 − u 2 v 2 Positive-definiteness (negative ho sakta hai).
C [ 0 , 1 ] ⟨ f , g ⟩ = ∫ 0 1 f ( x ) g ( x ) d x
Triangle inequality aur woh kis par depend karta hai ∥ u + v ∥ ≤ ∥ u ∥ + ∥ v ∥
Recall Feynman: 12-saal ke bachche ko explain karoSocho ek magic ruler hai. Normally ek ruler length measure karta hai. Lekin yeh magic ruler do arrows ke beech ka angle bhi bata sakta hai aur yeh bhi ki woh perfect corner bana rahe hain ya nahi. Dot product normal arrows ke liye wahi magic ruler hai. Ab socho tumhare "arrows" actually shapes hain, ya wiggly sound waves hain, ya numbers ki lists hain. Hume bas unke liye ek magic ruler invent karna hai — ek aisa rule jo in do cheezein leta ho aur ek number deta ho. Jab tak rule teen fairness rules follow kare (order se fark nahi padta, stretching ke saath theek se kaam karta hai, aur kisi cheez ki length kabhi negative nahi hoti), toh yeh turant shapes aur sounds ke liye lengths, angles, aur "right angles" de deta hai — chahe hum unhe draw bhi na kar sakein!
Mnemonic Teen axioms yaad karo:"SLP"
S ymmetry, L inearity, P ositive-definite → "S uper L egit P roduct" . Agar koi bhi letter fail ho, toh woh inner product nahi hai.
Dot Product — woh prototype jise yeh generalize karta hai.Norms and Distance — ⟨ v , v ⟩ Orthogonality and Gram-Schmidt — "perpendicular" define karne ke liye inner product chahiye.Cauchy-Schwarz Inequality — guarantee ki angles exist karte hain.Fourier Series — functions par integral inner product use karta hai.Orthogonal Projections — ⟨ ⋅ , ⋅ ⟩
Computation of projections
