Another Angle

Chapter 5 · Intuitive Math for Economics and Business

Directional Derivatives and the Gradient

A two-variable function hands every point on the floor a height, so its inputs fill a plane and its outputs trace a surface. Stand anywhere on it and you can step in any of three hundred and sixty degrees, each with its own slope. Measure just two of them, the one along x and the one along y, and every other direction is already decided.

The nudge ratio, one dimension up

Chapter 2 built the derivative as a nudge ratio: move the input a little, watch the output move, and compare the two. With one input there was nothing to decide. Forward or back was the whole choice, and the ratio came out as a single number.

On a surface that stops being true the moment you nudge. A floor is a plane, so "a little that way" needs a that way, and a direction with a size is a vector. That is why chapter 4 stopped to install the vocabulary. This chapter spends it.

Two directions come with names: hold y still and slide along x, and the rise is the partial derivative α; do it the other way and you get β. Glue each run and rise together and you get a tangent vector standing on the surface. The question the whole chapter is really asking is whether those two are enough for the other three hundred and fifty-eight.

They are, and the reason is the one idea from part one. Every direction on the floor is a combination of the two basic blocks, so its tangent vector is the same combination of the two tangent vectors. The whole floor follows those two up to a single tilted plane, and the rise in any direction reads off it as α cosθ + β sinθ. Written as a vector of the two partials dotted with the direction, that is Duf = ∇f · u. The gradient was never anything more exotic than those two numbers in a column.

The last third spends the formula rather than proving more of it: why the steepest climb is along the gradient, why implicit differentiation is a level curve refusing to change altitude, and why crowded contour lines mean steep ground. The final line, two gradients lining up where two curves touch, is the one the economics series leans on hardest.

Chapters in the video

Read it instead

The written version: implicit differentiation is a level curve's slope — where dy/dx = −fx/fy comes from, and why the minus sign was never arbitrary.

Walk inside it

The hillside shots in this chapter are recorded inside Open World Discover, and the ground you walk on there is the same surface as the one in the animation, fed straight into the terrain. Turning around really is sampling the slope in every direction.

← Chapter 4: the bare-minimum linear algebra
What's coming next →