\partial

Partial derivative

Operator: it does something

In plain English

A derivative with respect to one variable while all the others are frozen. When a quantity depends on many inputs, like a loss depending on millions of weights, $\frac{\partial L}{\partial w_3}$ asks how $L$ changes if you nudge only $w_3$ . The curly $\partial$ is a reminder that other variables exist and are being held fixed.

As code

// f depends on two inputs; vary only x
const f = (x, y) => x * x * y;
const h = 1e-6;
const dfdx = (x, y) => (f(x + h, y) - f(x, y)) / h; // y held fixed
dfdx(2, 5); // ≈ 20  (= 2xy)

See it

\partial L/\partial x

asks how the loss changes if you nudge only

x

and hold

y

fixed. The orange arrow is that one-direction slope, and the blue arrow is the same for

y

What it means, field by field

Physics

Field equations (∂/∂t, ∂/∂x); thermodynamics (∂U/∂S at fixed volume); the Schrödinger equation.

AI / ML

One component of the gradient: ∂L/∂θ_j. Backpropagation computes all of them efficiently with the chain rule.

Biology

Reaction-diffusion models of pattern formation (Turing 1952) are partial differential equations in time and space.

Mathematics

Multivariable calculus; also used for the boundary of a region (∂Ω), an unrelated meaning.

In a research paper

The time-dependent Schrödinger equation of quantum mechanics.

i\hbar\,\frac{\partial \psi}{\partial t} = \hat{H}\psi

How the wavefunction ψ changes in time (holding position fixed) is determined by the energy operator Ĥ acting on it.

Common misreadings

Treating ∂ and d as interchangeable. With many variables, ∂ signals that others are held constant, and which ones matters.
Reading ∂Ω (the boundary of a region) as a derivative. Same glyph, different job.

Connected symbols

\frac{d}{dx}

Derivative

\nabla

Gradient operator (nabla, del)

\psi

Psi

\mathcal{L}

Loss function (calligraphic L)

Where it appears on this site

§ 2. The Physics of Systems § 6. The Mathematics Inside Modern AI