Differentiation & Integration 📐

Calculus without the symbols. Just sums and differences.

Lecture 2-3

1. Why Numerical Calculus? 🤔

Slides 1-4

The Motivation

  • Black Box Functions: You can't analytically differentiate a simulation or experiment.
  • Complexity: Formulas like $\frac{d}{dx}(x \sin(x^2)\ln(x))$ are error-prone by hand.
  • Discrete Data: Real world data comes in points, not equations.
🤖

"Computers are bad at limits ($h \to 0$), but excellent at loops."

2. Numerical Differentiation

Slides 5-9

Forward Difference

$$ f'(x) \approx \frac{f(x+h) - f(x)}{h} $$

Looks simple, but suffers from the Step Size Dilemma. Too big = bad approximation. Too small = round-off noise.

3. Central Difference 🎯

Slides 10-13

Look both ways! Cancellations lead to better accuracy.

$$ f'(x) \approx \frac{f(x+h) - f(x-h)}{2h} $$ Error drops from $O(h)$ to $O(h^2)$.

4. Richardson Extrapolation 🚀

Slides 14-17

Combine two bad estimates to make one good one.

5. Numerical Integration 🧱

Slides 31-40

Trapezoidal Rule

Connect the dots with lines. Calculate area of trapezoids.

$$ \int_a^b f(x) dx \approx \frac{h}{2} [f(x_0) + 2f(x_1) + \dots + 2f(x_{n-1}) + f(x_n)] $$

5.1 Trapezoidal Implementation

Slides 31-40 (Cont.)

5.2 Simpson's Rule 🍩

Slides 31-40 (Cont.)

Simpson's Rule

Connect dots with parabolas. Weights: 1, 4, 1.

6. Advanced: Gaussian & SciPy 🧠

Slides 43-49

Gaussian Quadrature

Pick smart points $x_i$ and weights $w_i$. Integration becomes exact for polynomials.

SciPy `quad`

The industrial standard. Adaptive, fast, robust.