\int

Integral

Operator: it does something

In plain English

A sum over a continuum. The elongated S literally stands for "sum". $\int_a^b f(x)\,dx$ chops the interval from $a$ to $b$ into many thin slices of width $dx$ , multiplies each by the height $f(x)$ , and adds them, taking the limit as the slices get infinitely thin. Geometrically it is the (signed) area under the curve. Whenever a paper replaces $\sum$ with $\int$ , it is treating a quantity as continuous rather than discrete.

As code

// Riemann sum: an approximation that becomes exact as dx → 0
const f = (x) => x * x;
let area = 0, a = 0, b = 1, N = 100000;
const dx = (b - a) / N;
for (let i = 0; i < N; i++) area += f(a + i * dx) * dx;
// area ≈ 1/3

Where the analogy bends: The loop is a numerical approximation. The integral itself is the limit, and many integrals in physics are evaluated analytically rather than by looping.

See it

Number of slices21 · area ≈ 3.750

An integral is the limit of a Σ of thin rectangles,

\sum f(x_i)\,\Delta x \to \int f(x)\,dx

. More slices give a better area.

What it means, field by field

Physics

Work = ∫F·dx; total charge or mass from a density; the action ∫L dt; path integrals in quantum mechanics.

Mathematics

Area, accumulation, and the inverse of differentiation (Fundamental Theorem of Calculus).

AI / ML

Expectations over continuous distributions, 𝔼[f] = ∫ f(x) p(x) dx; marginalising latent variables in VAEs and diffusion models.

Biology

Total dose from a concentration-time curve (pharmacokinetic "AUC"); cumulative growth.

In a research paper

Probability theory; the training objectives of variational autoencoders and diffusion models are written this way.

\mathbb{E}_{x\sim p}[f(x)] = \int f(x)\,p(x)\,dx

The average value of f, weighting each possible x by how likely it is, over a continuous range of x.

Common misreadings

Thinking dx is decoration. It names the variable being summed over and carries units.
Assuming every integral can be solved in closed form. Most integrals in research are estimated numerically or by sampling.
Reading ∫ as "area" when the integrand is negative. Areas below the axis count negatively.

Connected symbols

\sum

Summation (capital sigma)

\frac{d}{dx}

Derivative

\mathbb{E}

Expectation (expected value)

\infty

Infinity