5.4.24 · D1 · Coding › Scientific Computing (Python) › Implementing numerical integration from scratch — trapezoida
Numerical integration ka matlab hai curve ke neeche ka area dhundhna — simple shapes (seedhi ya thodi-si curved tops waale boxes) se cover karke, jinke areas hum pehle se calculate karna jaante hain. Parent page par jo bhi hai woh sirf bookkeeping hai: curve ko kahan sample karo, kaun si shape use karo, aur final sum mein har sample kitna count karta hai.
Parent note par ek bhi formula padhne se pehle, tumhe uske har ek symbol par puri pakad chahiye. Neeche, har cheez ko zero se banaya gaya hai: pehle plain words mein, phir ek picture, phir topic ko yeh kyun chahiye . Upar se neeche padho — har item sirf wahi cheez use karta hai jo usse upar define ho chuki hai.
f ( x )
f ek ==rule hai jo x number leta hai aur f ( x ) number return karta hai==. x ko zameen par ek position samjho, aur f ( x ) ko uss jagah ke bilkul upar ek pahaadi ki height .
Picture. Horizontal axis ke saath left–right slide karo (yahi x choose karna hai); curve batata hai ki wahan pahaadi kitni oonchi hai.
Topic ko yeh kyun chahiye. Poora game is height curve ke neeche ka area measure karna hai. Height machine nahi → measure karne ke liye kuch nahi.
Definition Limits of integration
a left edge hai (jahan se measuring shuru hoti hai), b right edge hai (jahan khatam hoti hai). Square brackets [ a , b ] ka matlab hai "a se b tak ke saare x values, ends sammet".
Picture. x = a aur x = b par do seedhe fence-posts gade hue hain. Hum sirf dono fences ke beech pahaadi ke hisse ki parwah karte hain.
Topic ko yeh kyun chahiye. Region ke bina area ka koi matlab nahi. [ a , b ] zameen par wahi region hai .
Definition Integral ko zor se padhna
∫ a b f ( x ) d x
Padho: "f ke neeche ka area, x = a se x = b tak." Stretched-S ∫ purane zamaane ka S for Sum hai. d x ka matlab hai "ek sliver of width jo infinitely thin hai".
Picture. Fences ke beech aur curve ke neeche ke region ko paint se bharo. ∫ a b f d x paint ki matra hai — woh shaded area.
Topic ko yeh kyun chahiye. Yahi exact answer hai jisko hum dhundh rahe hain. Parent page par har method is ek quantity ka approximation hai.
Intuition "S for Sum" kyun?
Sochो ki shaded region ko vertical strips mein kaat rahe ho. Har strip hai (height f ( x ) ) × (tiny width d x ) = area ka ek sliver. Saare slivers jodon → total area. Integral wahi "infinitely many infinitely-thin slivers ko add karo" idea hai jo ek symbol mein freeze ho gaya hai. Yahi Riemann sums picture hai.
Hum infinitely many slivers handle nahi kar sakte, isliye finite number of chunks mein kaatte hain.
Definition Subintervals ki sankhya
n
n ==kitne equal pieces mein hum [ a , b ] ko kaatte hain==. Bada n = patle pieces = better approximation, zyada kaam.
x i
Cut points jahan pieces milte hain:
x i = a + i h , i = 0 , 1 , 2 , … , n .
Yahan i sirf ek counter hai (pehle ke liye 0, agle ke liye 1, …). x 0 = a left fence par hai, x n = b right fence par.
Picture. a se b tak rakha hua ek ruler, equal spacing h par ticked. Har tick ek node x i hai; tick ke neeche number uska index i hai.
Common mistake Points vs pieces — classic off-by-one
n pieces ke liye n + 1 ticks chahiye. (Do posts ek fence panel hold karte hain; teen posts do panels hold karte hain.) Isliye parent ka interior nodes ka loop range(1, n) hai — indices 1 se n − 1 tak.
Topic ko yeh kyun chahiye. h har area formula mein width factor hai; x i wahi jagahein hain jahan hum kabhi bhi f se height poochte hain.
Definition Subscript notation
x i ka matlab hai "i -th node". f ( x i ) (kabhi kabhi f i bhi likha jaata hai) ka matlab hai "uss node par curve ki height". Toh f 0 = f ( x 0 ) = f ( a ) aur f n = f ( x n ) = f ( b ) .
Picture. Har ruler tick par, curve tak ek vertical stick khींcho. Uski length f i hai. Ye sticks wahi numbers hain jo methods kabhi bhi plug in karte hain.
Topic ko yeh kyun chahiye. Dono rules sirf f 0 , f 1 , … , f n numbers se bani recipes hain — aur kuch nahi.
Definition Summation symbol
∑ i = 1 n − 1 f ( x i ) = f ( x 1 ) + f ( x 2 ) + ⋯ + f ( x n − 1 ) .
Greek capital sigma Σ ka matlab sum hai. Neeche (i = 1 ) batata hai counter kahan shuru hota hai; upar (n − 1 ) batata hai kahan rukta hai; body (f ( x i ) ) woh hai jo har baar add karna hai.
Picture. Ek conveyor belt: yeh i = 1 , 2 , … , n − 1 ek-ek karke feed karta hai, har baar f ( x i ) ek running total mein dalta hai.
Topic ko yeh kyun chahiye. Trapezoidal ya Simpson ka answer ek weighted sum ∑ i w i f ( x i ) hai. Sigma compact tarika hai "saari sampled heights ko add karo" likhne ka.
Weight w i ek multiplier hai jo batata hai ki ek node ki height total mein kitni baar count hoti hai . Interior nodes neighbouring shapes ke beech share hote hain, isliye akele endpoints se zyada count karte hain.
Intuition Weights exist kyun karte hain
Do adjacent shapes ki boundary par baithe ek node ka belonging dono shapes se hai. Jab tum shapes ke areas jodte ho, woh shared height do baar add hoti hai — weight 2 (trapezoid) ya 2 /4 (Simpson). Ek endpoint sirf ek shape ko touch karta hai, isliye ek baar count hota hai. Trapezoidal aur Simpson's ke beech ka poora fark yahi hai ki kaunse weights woh use karte hain .
Picture. Har node ko is hisaab se colour karo ki uspar kitni shapes lean karti hain: endpoints (ek shape) ek colour, shared interior nodes (do shapes) doosra colour.
Topic ko yeh kyun chahiye. Endpoints ko galat weight dena parent note ka "#1 silent bug" hai. Kyun weights alag hain yeh samajhna us bug ko impossible bana deta hai.
Picture. Line = ruler ka kinara. Parabola = ek smooth valley ya hill. Cubic = ek gentle S.
Topic ko yeh kyun chahiye. Trapezoid har slice ko ek line (degree 1) fit karta hai; Simpson har pair ko ek parabola (degree 2). Parent note par "degree-1 / degree-3 ke liye exact" ek statement hai ki kaunsi curves yeh shapes perfectly reproduce karti hain . Dekho Polynomial interpolation aur Newton-Cotes formulas .
Definition Order of error
O ( h 2 ) ka matlab hai "error roughly h 2 ke proportional hai". Agar tum h half karo, toh O ( h 2 ) error lagbhag quarter ho jaati hai; O ( h 4 ) error lagbhag sixteenth ho jaati hai.
Picture. Do shrinking staircases: h half karna O ( h 2 ) error ko ek chhota stairs neeche le jaata hai, lekin O ( h 4 ) error ko chaar stairs neeche phenk deta hai. Same effort, Simpson jeet jaata hai.
Topic ko yeh kyun chahiye. Yeh parent ki punchline explain karta hai: same number of samples ke liye Simpson trapezoid ko dramatically beat karta hai. Powers khud Taylor series se aate hain.
≈
≈ ka matlab hai "close to, lekin exactly nahi" . Hum ∫ a b f d x ≈ T n likhte hain kyunki shapes sirf curve ko mimic karti hain; woh ise perfectly reproduce nahi karti (jab tak curve ittefaqan low-degree polynomial na ho).
Topic ko yeh kyun chahiye. Honesty. Har method ek estimate hai; ≈ ise front aur centre rakhta hai.
Function f x height machine
Weights wi shared vs edge
Polynomial degree line vs parabola
Jab upar ka har box obvious lagne lage, parent page — the main topic — sirf pure bookkeeping ki tarah padhega.
Khud ko test karo: left side padho, answer zor se bolo, phir reveal karo.
∫ a b f ( x ) d x ka plain words mein kya matlab hai?x = a aur x = b ke beech curve f ke neeche ka exact area.
h kya hai aur yeh kaise compute hota hai?Har subinterval ki width; h = ( b − a ) / n .
Agar n subintervals ki sankhya hai, toh kitne nodes (points) hain? n + 1 .
Node formula x i = a + ih tumhe kya deta hai? [ a , b ] ke saath i -th cut point ki position.
∑ i = 1 n − 1 f ( x i ) ko zor se padho.x 1 se x n − 1 tak har interior node par curve ki heights ka sum.
Interior nodes ko endpoints se bada weight kyun milta hai? Woh do neighbouring shapes ke beech share hote hain, isliye unki height dono mein count hoti hai.
Trapezoid har slice ke liye kaun si shape (degree) fit karta hai? Simpson har pair ke liye? Trapezoid: seedhi line (degree 1). Simpson: parabola (degree 2).
Agar tum h half karo, toh O ( h 2 ) error kitna shrink hoti hai? O ( h 4 ) error? Lagbhag quarter ho jaati hai; lagbhag sixteenth ho jaati hai.
∫ f d x ≈ T n mein ≈ kya signal karta hai?Result ek approximation hai, exact nahi, jab tak f low-degree polynomial na ho.
Interior-node loop range(1, n) use kyun karta hai, range(1, n+1) nahi? Indices 1 se n − 1 tak interior nodes hain; x n endpoint hai, alag handle hota hai.