ContentsThe library

Matrices as Maps

When You Cannot Hit the Target

Last timeThe Stretch in Each Direction

Most real systems ask for something the map cannot produce, so the honest answer is the closest reachable point, and that one idea is the whole of fitting.

Suppose you have forty measurements and three unknowns. The map from unknowns to

predicted measurements has three columns, so its reachable set is a flat piece of

at most three directions sitting inside a forty dimensional space. The target, your

actual measurements, sits somewhere in that forty dimensional space, and the chance

that it happens to lie exactly on a three dimensional flat piece is nil.

So the system has no solution. By the previous lessons this is the expected answer

rather than a problem: the target is not in the reachable set. What makes it useful

is that the question can be changed. Instead of asking which input produces the

target exactly, ask which input produces the closest thing to the target that the

map can produce at all.

That question always has an answer, and finding it is the single most used piece of

linear algebra in practice.

Drop a perpendicular

The geometry is the kind of thing a child knows. Given a plane and a point above

it, the nearest point of the plane is directly below, found by dropping a

perpendicular. Any other point of the plane is further away, and the reason is the

right triangle: the line from the outside point to the other candidate is a

hypotenuse, with the perpendicular drop as one leg, and a hypotenuse is always

longer than a leg.

FIG 1The nearest reachable point
Everything about fitting is in this picture. The reachable set and the destroyed set from the fourth lesson are both here: the reachable set is what you project onto, and if anything is destroyed then several inputs produce the same closest point and the best fit is not unique.

Where the formula comes from

Now transcribe the geometry. The leftover is perpendicular to the reachable set,

and the reachable set is spanned by the columns, so the leftover is perpendicular

to every column. That is one equation per column, which is one equation per

unknown, which is a square system.

The lesson stops here

4 more paragraphs to go

You have read the opening. The rest of the argument, the problems that check whether it landed, and the lines worth keeping at the end all come with a plan.

The first lesson of every course in the library reads the whole way through, free, so you can see exactly what the rest of them are.

See the planThe contents

This is the reading half

Starting the course gives you your own copy of it. Every idea on every page has problems standing under it, marked with a reason rather than a tick, and any sentence you do not believe can be opened and argued with. None of that can happen on a page nobody owns.

The contents

The rest of this course

  1. 01Not a Grid, a Verb
  2. 02Each Column Is a Destinationopening only
  3. 03Why the Multiplication Rule Looks Like Thatopening only
  4. 04The Shadow and the Thingopening only
  5. 05Undo, If You Canopening only
  6. 06The Directions That Only Stretchopening only
  7. 07Every Map Is a Rotation, a Stretch and a Rotationopening only
  8. 08When You Cannot Hit the Targetyou are here

Read alongside