Boo-AI — Master Artificial Intelligence by Building from Scratch

Learning Objectives

By the end of this section, you will be able to:

Understand the Jacobian matrix as the derivative of a multivariable transformation
Compute the Jacobian determinant for 2D and 3D coordinate transformations
Interpret the Jacobian geometrically as a local area/volume scaling factor
Apply the change of variables formula to evaluate double and triple integrals
Derive the Jacobians for polar, cylindrical, and spherical coordinates
Connect the Jacobian to machine learning applications including normalizing flows and probability transformations

The Big Picture: Transforming Integrals

"The Jacobian is the multivariable generalization of the chain rule's factor that appears in u-substitution. It tells us how areas and volumes stretch and compress under coordinate transformations."

In single-variable calculus, u-substitution lets us transform an integral to a new variable: $\int f(g(x))g'(x)\,dx = \int f(u)\,du$ . The derivative $g'(x)$ appears as a correction factor that accounts for how the variable change stretches or compresses intervals.

In multiple dimensions, the analogous concept is the Jacobian determinant. When we transform from one coordinate system $(u, v)$ to another $(x, y)$ , the Jacobian tells us how infinitesimal area elements transform: $dA_{xy} = |J| \, dA_{uv}$ .

Why This Matters

Change of variables is essential for:

Simplifying integrals: Converting complex regions to rectangles or other simple shapes
Exploiting symmetry: Using polar coordinates for circular regions, spherical for spheres
Physics: Transforming between coordinate systems natural to different problems
Machine learning: Normalizing flows, probability transformations, and generative models
Computer graphics: Texture mapping, deformations, and coordinate transforms

Historical Context

The Jacobian is named after Carl Gustav Jacob Jacobi (1804–1851), a German mathematician who made fundamental contributions to analysis, number theory, and mechanics. Jacobi systematically studied determinants of partial derivatives in the 1830s, developing the theory we use today.

The underlying ideas, however, date back to the work of Lagrange and Laplace on celestial mechanics, where transformations between coordinate systems were essential for computing gravitational interactions. The change of variables formula for multiple integrals was developed rigorously in the 19th century by mathematicians including Augustin-Louis Cauchy and Bernhard Riemann.

Jacobi's Legacy

Beyond the Jacobian, Carl Jacobi contributed to elliptic functions, the theory of determinants, and Hamiltonian mechanics. His work on canonical transformations in mechanics directly uses Jacobian determinants to ensure that phase space volume is preserved—a result now known as Liouville's theorem.

Why Change Variables?

Consider computing $\iint_R e^{-(x^2 + y^2)} \, dA$ over a disk of radius $a$ centered at the origin. In Cartesian coordinates, this requires integrating over the region $-a \le x \le a$ , $-\sqrt{a^2 - x^2} \le y \le \sqrt{a^2 - x^2}$ —a mess of square roots.

In polar coordinates, the same region is simply $0 \le r \le a$ , $0 \le \theta \le 2\pi$ —a rectangle! And the integrand $e^{-(x^2 + y^2)}$ becomes $e^{-r^2}$ , which depends only on $r$ .

But we cannot simply write $\int_0^{2\pi} \int_0^a e^{-r^2} \, dr \, d\theta$ . This would give the wrong answer! The area element $dx \, dy$ is not equal to $dr \, d\theta$ . We need a correction factor—the Jacobian—that accounts for how the transformation stretches area elements.

The Fundamental Question

If $x = x(u, v)$ and $y = y(u, v)$ , then

$dx \, dy = \text{?} \cdot du \, dv$

What factor relates the area elements in different coordinate systems?

The Jacobian Matrix

Let $T: \mathbb{R}^2 \to \mathbb{R}^2$ be a transformation defined by $x = x(u, v)$ and $y = y(u, v)$ . The Jacobian matrix of $T$ is the matrix of all first-order partial derivatives:

Definition: Jacobian Matrix

J = \frac{\partial(x, y)}{\partial(u, v)} = \begin{pmatrix} \frac{\partial x}{\partial u} & \frac{\partial x}{\partial v} \\ \frac{\partial y}{\partial u} & \frac{\partial y}{\partial v} \end{pmatrix}

The Jacobian matrix is the total derivative of the transformation—the best linear approximation to $T$ near a point. It tells us how small changes in $(u, v)$ affect changes in $(x, y)$ :

\begin{pmatrix} dx \\ dy \end{pmatrix} \approx J \begin{pmatrix} du \\ dv \end{pmatrix}

Example: Polar Coordinates

For polar coordinates with $x = r\cos\theta$ and $y = r\sin\theta$ :

\frac{\partial x}{\partial r} = \cos\theta

\frac{\partial x}{\partial \theta} = -r\sin\theta

\frac{\partial y}{\partial r} = \sin\theta

\frac{\partial y}{\partial \theta} = r\cos\theta

J = \begin{pmatrix} \cos\theta & -r\sin\theta \\ \sin\theta & r\cos\theta \end{pmatrix}

The Jacobian Determinant

For changing variables in integrals, we need the determinant of the Jacobian matrix, often called simply "the Jacobian":

Jacobian Determinant (2D)

|J| = \det(J) = \frac{\partial x}{\partial u} \cdot \frac{\partial y}{\partial v} - \frac{\partial x}{\partial v} \cdot \frac{\partial y}{\partial u}

Computing for Polar Coordinates

|J| = \det \begin{pmatrix} \cos\theta & -r\sin\theta \\ \sin\theta & r\cos\theta \end{pmatrix}

= \cos\theta \cdot r\cos\theta - (-r\sin\theta) \cdot \sin\theta

= r\cos^2\theta + r\sin^2\theta

|J| = r

This is exactly the factor we need: $dA = r \, dr \, d\theta$ in polar coordinates!

Geometric Interpretation

The Jacobian determinant has a beautiful geometric meaning: it measures how the transformation scales areas (in 2D) or volumes (in 3D).

Key Geometric Insight

Consider a small rectangle in $(u, v)$ space with sides $\Delta u$ and $\Delta v$ . Under the transformation, this rectangle maps to a (curved) parallelogram in $(x, y)$ space.

The area of the original rectangle is $\Delta u \cdot \Delta v$ . The area of the transformed parallelogram is approximately $|J| \cdot \Delta u \cdot \Delta v$ .

\text{Transformed Area} \approx |J| \times \text{Original Area}

The two sides of the rectangle transform to the vectors formed by the columns (or rows) of the Jacobian matrix. The area of the parallelogram spanned by two vectors is the absolute value of their cross product (in 2D, this is the determinant).

Signed vs. Unsigned

The determinant $\det(J)$ can be positive or negative. A negative determinant indicates the transformation reverses orientation (like a reflection). For area calculations, we use the absolute value $|J|$ .

Interactive: Area Scaling Under Transformation

Explore how a rectangular element in parameter space transforms to a curved quadrilateral in physical space. Move the highlighted region and observe how the Jacobian varies with position:

🔲Area Scaling Under Polar Transformation

r (radius): 1.00

θ (angle): 45°

Δ (size): 0.30

(r, θ) Parameter Space

Area = Δr × Δθ = 0.0450

(x, y) Physical Space

Area ≈ |J| × Δr × Δθ = 0.0517

\frac{\partial \mathbf{r}}{\partial r} \cdot \Delta r

\frac{\partial \mathbf{r}}{\partial \theta} \cdot \Delta \theta

Area Scaling Analysis

Parameter Area

0.0450

Jacobian |J| = r

1.1500

Physical Area

0.0517

dA = r \, dr \, d\theta

The area element in polar coordinates picks up a factor of r from the Jacobian

Key Insight: Notice how the transformed region (green) is larger when r is larger. This is because the Jacobian |J| = r scales the area element. At the origin (r = 0), the area element would shrink to zero, which is why polar coordinates are singular there.

The Change of Variables Formula

Now we can state the main theorem that makes it all work:

Change of Variables Theorem (2D)

Let $T: S \to R$ be a one-to-one transformation with continuous partial derivatives. If $f$ is continuous on $R$ , then:

\iint_R f(x, y) \, dA = \iint_S f(x(u,v), y(u,v)) \left| \frac{\partial(x,y)}{\partial(u,v)} \right| \, du \, dv

In words: to change from $(x, y)$ to $(u, v)$ coordinates:

Replace $x$ and $y$ with their expressions in $u$ and $v$
Multiply by the absolute value of the Jacobian $|J|$
Replace $dA$ with $du \, dv$
Change the limits of integration to describe the same region in $(u, v)$ coordinates

Interactive: Jacobian Transform Visualizer

Explore how different coordinate transformations distort space. The highlighted cell shows the local area scaling factor (the Jacobian):

📐Interactive Jacobian Transform Visualizer

Transformation Type

Grid Density: 6

Transform from (r, θ) to (x, y). The Jacobian |J| = r means area elements scale by r.

(r, θ) Space

Area = 0.3491

→

Transform

|J| = 0.5000

(x, y) Space

Area ≈ 0.1745

Highlight r: 0.50

Highlight θ: 0.50

Jacobian Determinant

\frac{\partial(x,y)}{\partial(r,\theta)} = \begin{vmatrix} \cos\theta & -r\sin\theta \\ \sin\theta & r\cos\theta \end{vmatrix} = r

The green region's area scales by |J| = 0.5000 at this position

Things to Explore

In polar coordinates, observe how the Jacobian = r means cells near the origin have smaller physical area
For linear transformations, the Jacobian is constant everywhere
Try the parabolic and elliptical coordinates to see non-uniform scaling

Example: Polar Coordinates in Detail

Let's work through a complete example using polar coordinates.

Setup

The transformation from polar $(r, \theta)$ to Cartesian $(x, y)$ is:

x = r\cos\theta, \quad y = r\sin\theta

We computed $|J| = r$ , so:

dA = dx \, dy = r \, dr \, d\theta

Example: Area of a Disk

Find the area of a disk of radius $a$ :

A = \iint_R 1 \, dA = \int_0^{2\pi} \int_0^a r \, dr \, d\theta

Inner integral:

\int_0^a r \, dr = \frac{r^2}{2} \Big|_0^a = \frac{a^2}{2}

Outer integral:

\int_0^{2\pi} \frac{a^2}{2} \, d\theta = \frac{a^2}{2} \cdot 2\pi = \pi a^2

Example: Gaussian Integral

Evaluate $\iint_{\mathbb{R}^2} e^{-(x^2+y^2)} \, dA$ :

= \int_0^{2\pi} \int_0^\infty e^{-r^2} \cdot r \, dr \, d\theta

Inner integral (u-substitution with u = r²):

\int_0^\infty r e^{-r^2} \, dr = \frac{1}{2} \int_0^\infty e^{-u} \, du = \frac{1}{2}

Result:

= 2\pi \cdot \frac{1}{2} = \pi

Connection to Probability

This result implies $\int_{-\infty}^{\infty} e^{-x^2} dx = \sqrt{\pi}$ , which normalizes the Gaussian distribution. The change to polar coordinates is the classic trick for evaluating this integral!

Cylindrical and Spherical Coordinates

Cylindrical Coordinates

$x = r\cos\theta, \quad y = r\sin\theta, \quad z = z$

The Jacobian for cylindrical is the same as polar in the xy-plane, with z contributing a factor of 1:

dV = r \, dr \, d\theta \, dz

Spherical Coordinates

$x = \rho\sin\phi\cos\theta, \quad y = \rho\sin\phi\sin\theta, \quad z = \rho\cos\phi$

The 3×3 Jacobian matrix has determinant:

|J| = \rho^2 \sin\phi

dV = \rho^2 \sin\phi \, d\rho \, d\phi \, d\theta

Coordinate System	Jacobian \|J\|	Volume Element dV
Cartesian	1	dx dy dz
Polar (2D)	r	r dr dθ
Cylindrical	r	r dr dθ dz
Spherical	ρ² sin φ	ρ² sin φ dρ dφ dθ

Interactive: Jacobian Step-by-Step Calculator

Follow the step-by-step process of computing the Jacobian determinant for various coordinate transformations:

🧮Jacobian Step-by-Step Calculator

Select Transformation

Step 1: Identify the Transformation

We have a transformation from (r, \theta) to (x, y):

x = r \cos\theta

y = r \sin\theta

Worked Examples

Example 1: Elliptical Region

Evaluate $\iint_R (x^2 + y^2) \, dA$ where R is the region bounded by $\frac{x^2}{4} + \frac{y^2}{9} = 1$ .

Solution: Use the transformation $x = 2u, y = 3v$ . Then $u^2 + v^2 = 1$ is the unit circle.

Jacobian: $|J| = \left|\begin{matrix} 2 & 0 \\ 0 & 3 \end{matrix}\right| = 6$

= \iint_{u^2+v^2 \le 1} (4u^2 + 9v^2) \cdot 6 \, du \, dv

Converting to polar for the unit disk:

= 6 \int_0^{2\pi} \int_0^1 (4r^2\cos^2\theta + 9r^2\sin^2\theta) \cdot r \, dr \, d\theta

After integration: $= 6 \cdot \frac{1}{4} \cdot \frac{1}{2}(4 + 9) \cdot 2\pi = \frac{39\pi}{2}$

Example 2: Volume of a Sphere

Find the volume of a sphere of radius $a$ .

Solution: Use spherical coordinates:

V = \int_0^{2\pi} \int_0^\pi \int_0^a \rho^2 \sin\phi \, d\rho \, d\phi \, d\theta

Inner integral:

\int_0^a \rho^2 \, d\rho = \frac{a^3}{3}

Middle integral:

\int_0^\pi \sin\phi \, d\phi = 2

Outer integral:

V = \frac{a^3}{3} \cdot 2 \cdot 2\pi = \frac{4\pi a^3}{3}

Applications

Physics: Computing Moments of Inertia

The moment of inertia of a solid about an axis often simplifies in cylindrical or spherical coordinates. For a sphere of uniform density about a diameter:

I = \iiint_E r^2 \rho \, dV = \int_0^{2\pi} \int_0^\pi \int_0^a (\rho^2\sin^2\phi) \cdot \rho_0 \cdot \rho^2\sin\phi \, d\rho \, d\phi \, d\theta

Probability: Multivariate Distributions

When transforming random variables, the Jacobian adjusts the probability density. If $Y = g(X)$ where $g$ is invertible:

f_Y(y) = f_X(g^{-1}(y)) \cdot \left| \det\left(\frac{\partial g^{-1}}{\partial y}\right) \right|

Machine Learning Connection

The Jacobian is central to several machine learning techniques:

Normalizing Flows

Normalizing flows are generative models that transform a simple base distribution (like a Gaussian) through a series of invertible transformations to produce a complex target distribution.

The log-likelihood of a sample involves the Jacobian:

\log p(x) = \log p(z) - \sum_{i=1}^{K} \log \left| \det\left(\frac{\partial f_i}{\partial z_{i-1}}\right) \right|

Each transformation $f_i$ must have a tractable Jacobian determinant for efficient training.

Variational Autoencoders (VAEs)

The reparameterization trick in VAEs uses the Jacobian. When sampling $z = \mu + \sigma \cdot \epsilon$ where $\epsilon \sim N(0, 1)$ , the Jacobian is $\sigma$ , which appears in the density transformation.

Independent Component Analysis (ICA)

ICA finds a linear transformation to independent components. The log-likelihood includes a log-Jacobian term that accounts for the mixing matrix's effect on volume.

Python Implementation

Computing Jacobians

Computing Jacobian Matrices and Determinants

🐍jacobian_computation.py

Explanation(4)

Code(41)

5Jacobian Matrix Definition

The Jacobian matrix contains all first-order partial derivatives of the transformation. For 2D, it's a 2×2 matrix.

11Symbolic Differentiation

We use SymPy to compute partial derivatives symbolically, then evaluate at specific points for numerical accuracy.

18Matrix Construction

The Jacobian matrix J has entries: J₁₁ = ∂x/∂u, J₁₂ = ∂x/∂v, J₂₁ = ∂y/∂u, J₂₂ = ∂y/∂v.

29Polar Coordinates

For polar: x = r cos θ, y = r sin θ. The Jacobian determinant equals r, as expected.

37 lines without explanation

1import numpy as np
2from scipy import integrate
3import sympy as sp
4
5def jacobian_matrix(x_funcs, y_funcs, u, v, u_val, v_val):
6    """
7    Compute the Jacobian matrix numerically at a point.
8
9    x_funcs, y_funcs: expressions for x(u,v), y(u,v)
10    u, v: SymPy symbols
11    u_val, v_val: point to evaluate at
12    """
13    # Compute partial derivatives symbolically
14    dxdu = sp.diff(x_funcs, u)
15    dxdv = sp.diff(x_funcs, v)
16    dydu = sp.diff(y_funcs, u)
17    dydv = sp.diff(y_funcs, v)
18
19    # Evaluate at the point
20    J = np.array([
21        [float(dxdu.subs([(u, u_val), (v, v_val)])),
22         float(dxdv.subs([(u, u_val), (v, v_val)]))],
23        [float(dydu.subs([(u, u_val), (v, v_val)])),
24         float(dydv.subs([(u, u_val), (v, v_val)]))]
25    ])
26
27    return J
28
29# Example: Polar coordinates
30u, v = sp.symbols('r theta', real=True)
31x = u * sp.cos(v)  # x = r*cos(theta)
32y = u * sp.sin(v)  # y = r*sin(theta)
33
34# Evaluate Jacobian at r=2, theta=pi/4
35r_val, theta_val = 2, np.pi/4
36J = jacobian_matrix(x, y, u, v, r_val, theta_val)
37
38print("Jacobian Matrix at (r=2, θ=π/4):")
39print(J)
40print(f"\nJacobian Determinant |J| = {np.linalg.det(J):.4f}")
41print(f"Expected value (r = 2): {r_val}")

Integration with Change of Variables

Polar Coordinate Integration

🐍polar_integration.py

Explanation(4)

Code(41)

5Polar Integration Function

This function automates the change of variables to polar coordinates, including the Jacobian factor r.

12The Key Step: Jacobian

We multiply f(x, y) by r, the Jacobian for polar coordinates. This is the crucial factor that makes the change of variables formula work.

24Gaussian Integral

The integral of e^(−r²) over a disk is much easier in polar: the integrand becomes r·e^(−r²), which has antiderivative −½e^(−r²).

32Radial Function

When f depends only on r = √(x² + y²), polar coordinates are ideal. Here f = r, so the integrand is r · r = r².

37 lines without explanation

1import numpy as np
2from scipy import integrate
3
4def polar_integral(f, r_range, theta_range):
5    """
6    Compute ∬_R f(x,y) dA using polar coordinates.
7
8    Uses the change of variables formula:
9    ∬_R f(x,y) dA = ∬_S f(r cos θ, r sin θ) · r dr dθ
10    """
11    def integrand(theta, r):
12        x = r * np.cos(theta)
13        y = r * np.sin(theta)
14        return f(x, y) * r  # Multiply by Jacobian |J| = r
15
16    result, error = integrate.dblquad(
17        integrand,
18        r_range[0], r_range[1],  # r limits
19        lambda r: theta_range[0],  # theta lower
20        lambda r: theta_range[1]   # theta upper
21    )
22    return result
23
24# Example 1: ∬ e^(-(x² + y²)) dA over disk of radius 2
25f1 = lambda x, y: np.exp(-(x**2 + y**2))
26result1 = polar_integral(f1, (0, 2), (0, 2*np.pi))
27print(f"∬ e^(-(x² + y²)) dA over disk R=2: {result1:.6f}")
28print(f"Exact value π(1 - e^(-4)): {np.pi * (1 - np.exp(-4)):.6f}")
29
30# Example 2: ∬ √(x² + y²) dA over quarter disk
31f2 = lambda x, y: np.sqrt(x**2 + y**2)
32result2 = polar_integral(f2, (0, 1), (0, np.pi/2))
33print(f"\n∬ √(x² + y²) dA over quarter disk: {result2:.6f}")
34print(f"Exact value π/6: {np.pi/6:.6f}")
35
36# Example 3: Volume of hemisphere
37# z = √(1 - x² - y²) over unit disk
38f3 = lambda x, y: np.sqrt(max(0, 1 - x**2 - y**2))
39result3 = polar_integral(f3, (0, 1), (0, 2*np.pi))
40print(f"\nVolume of hemisphere (half of 4π/3 · 1³ / 2): {result3:.6f}")
41print(f"Exact value 2π/3: {2*np.pi/3:.6f}")

Normalizing Flows Example

Jacobian in Normalizing Flows

🐍normalizing_flow.py

Explanation(4)

Code(51)

5Normalizing Flows

Normalizing flows are generative models that transform simple distributions to complex ones using invertible maps with tractable Jacobians.

22Log Jacobian

In practice, we work with log probabilities and log Jacobians for numerical stability. The log-det-Jacobian adjusts probability densities under transformation.

30Density Transformation

The change of variables formula for probability: p(x) = p(z) / |det(∂x/∂z)|. Taking logs: log p(x) = log p(z) - log |det J|.

38ML Application

This is the core mechanism behind normalizing flows, VAEs with flow posteriors, and density estimation networks.

47 lines without explanation

1import numpy as np
2
3class NormalizingFlowTransform:
4    """
5    A simple normalizing flow transformation demonstrating
6    the role of the Jacobian in probability transformations.
7    """
8
9    def __init__(self, scale, shift):
10        self.scale = scale  # Scaling factor
11        self.shift = shift  # Translation
12
13    def forward(self, z):
14        """Transform z to x."""
15        return self.scale * z + self.shift
16
17    def inverse(self, x):
18        """Transform x back to z."""
19        return (x - self.shift) / self.scale
20
21    def log_det_jacobian(self, z):
22        """
23        Log of the absolute Jacobian determinant.
24        For affine transform: |J| = |scale|
25        """
26        return np.log(np.abs(self.scale))
27
28    def transform_density(self, z, base_log_prob):
29        """
30        Transform a sample and compute the new log probability.
31
32        p(x) = p(z) / |det(dx/dz)|
33        log p(x) = log p(z) - log|det(J)|
34        """
35        x = self.forward(z)
36        log_prob = base_log_prob - self.log_det_jacobian(z)
37        return x, log_prob
38
39# Example: Transform standard normal to N(5, 2²)
40flow = NormalizingFlowTransform(scale=2.0, shift=5.0)
41
42# Sample from standard normal
43z_samples = np.random.randn(5)
44base_log_probs = -0.5 * z_samples**2 - 0.5*np.log(2*np.pi)
45
46print("Normalizing Flow Example: N(0,1) → N(5, 4)")
47print("=" * 50)
48for i, (z, base_lp) in enumerate(zip(z_samples, base_log_probs)):
49    x, new_lp = flow.transform_density(z, base_lp)
50    print(f"z = {z:7.4f} → x = {x:7.4f}")
51    print(f"  log p(z) = {base_lp:7.4f} → log p(x) = {new_lp:7.4f}")

Common Mistakes

Mistake 1: Forgetting the Jacobian

The most common error is writing $\iint f(r\cos\theta, r\sin\theta) \, dr \, d\theta$ without the factor of $r$ . Always include $|J|$ !

Mistake 2: Wrong Limits

When changing variables, you must transform the limits too. A disk of radius 2 in Cartesian becomes $0 \le r \le 2$ , $0 \le \theta \le 2\pi$ —not the original x and y limits!

Mistake 3: Signed vs. Unsigned Jacobian

For area/volume calculations, use $|J|$ (absolute value). The signed determinant matters for orientation, but area is always positive.

Pro Tip: Check Your Answer

When learning, verify with a known result. Compute the area of a circle using polar coordinates—you should get $\pi r^2$ . If not, you likely forgot the Jacobian or set up limits incorrectly.

Test Your Understanding

❓Jacobian QuizQuestion 1 of 8

For polar coordinates where

x = r\cos\theta

and

y = r\sin\theta

, what is the Jacobian determinant

\left|\frac{\partial(x,y)}{\partial(r,\theta)}\right|

Current score: 0 / 0

Summary

The Jacobian is the key to changing variables in multiple integrals. It captures how a transformation stretches or compresses area and volume elements.

Key Formulas

Concept	Formula
Jacobian Matrix (2D)	J = [∂x/∂u ∂x/∂v; ∂y/∂u ∂y/∂v]
Jacobian Determinant	\|J\| = ∂x/∂u · ∂y/∂v − ∂x/∂v · ∂y/∂u
Change of Variables (2D)	∬_R f(x,y) dA = ∬_S f(x(u,v), y(u,v)) \|J\| du dv
Polar	\|J\| = r, dA = r dr dθ
Cylindrical	\|J\| = r, dV = r dr dθ dz
Spherical	\|J\| = ρ² sin φ, dV = ρ² sin φ dρ dφ dθ

Key Takeaways

The Jacobian matrix is the matrix of all first-order partial derivatives—the total derivative of the transformation.
The Jacobian determinant measures local area/volume scaling—how much an infinitesimal element stretches or shrinks.
Change of variables requires multiplying by |J| to account for this scaling.
Polar: |J| = r; Cylindrical: |J| = r; Spherical: |J| = ρ² sin φ. These are the most common cases.
In machine learning, Jacobians appear in normalizing flows, VAEs, and probability transformations.
Always remember: transform both the integrand AND the limits, and include the Jacobian factor.

The Essence of the Jacobian:

"The Jacobian determinant tells us how much a transformation stretches or compresses space—the scaling factor that makes change of variables work."

What's Next: With the Jacobian, you can now tackle integrals in any coordinate system. In subsequent chapters, we'll explore vector calculus, where the Jacobian appears in transformations of vector fields and in the proofs of Green's, Stokes', and the Divergence theorems.