Latest

You can't just optimize freely when constraints are in play. Lagrange multipliers turn a constrained optimization into a system of equations by asking: at the solution, which way do the gradients point?

When you integrate ∫∫ r dr dθ in polar coordinates instead of ∫∫ dx dy in Cartesian, where does that extra r factor come from? It's the Jacobian—the scaling factor that accounts for how area (or volume) elements stretch or compress when you change coordinate systems. This is the

A double integral sweeps area; a triple integral sweeps volume. Whether you're calculating the total mass of a non-uniform solid or the total charge inside a region of space, the technique is the same: integrate over all three spatial variables, in the right order.

One integral finds area under a curve. Two integrals do the same trick over a surface. Double integrals are the tool that makes spatial calculus click.

The single-variable chain rule says how to differentiate a composition of functions. In multiple dimensions, the same idea produces a sum of partial derivative products — and that sum is exactly what backpropagation computes at every layer of a neural network.

The partial derivative only measures change along coordinate axes. The directional derivative asks: what's the rate of change if you move in any arbitrary direction? The answer reveals the gradient's true geometry.

The gradient is the multivariable calculus equivalent of the derivative—but it's not a number. It's a vector that points in the direction where the function increases most rapidly, with magnitude equal to the rate of that increase. This is geometric thinking elevated to computational power.