The Row and Column Pictures

One system, two pictures. Read by rows, $A x = b$ asks where several flats intersect. Read by columns, it asks whether $b$ can be mixed from the columns of $A$ . Both describe the same solutions, and each makes visible what the other hides, which is why moving between them is a competence rather than a preference.

Definition

In the row picture, each equation $a_{i}^{T} x = b_{i}$ is a hyperplane in $R^{n}$ , and the solution set of $A x = b$ is the intersection of all $m$ of them. In the column picture, the same system asks for weights $x_{1}, \dots, x_{n}$ such that $x_{1} A_{1} + \dots + x_{n} A_{n} = b$ , that is, whether $b$ lies in the span of the columns of $A$ . The two are the same question: one asks where flats meet, the other asks what can be reached.

Formal statement

Row: $⋂_{i = 1}^{m} {x : a_{i}^{T} x = b_{i}}$ . Column: $b \in span {A_{1}, \dots, A_{n}}$ , that is $b = \sum_{j} x_{j} A_{j}$ .

Assumptions and scope

The two pictures describe the same solution set. Any conclusion reached in one must hold in the other, and a disagreement between them is an error rather than a discovery.
The row picture lives in $R^{n}$ , the space of the unknowns; the column picture lives in $R^{m}$ , the space of the right-hand side. For a non-square system the two diagrams are drawn in different spaces.
Parallel rows correspond to dependent columns only in the square case. In general the correspondence runs through rank, not through a visual analogy.
The column picture answers the existence question directly, because it asks whether $b$ is reachable at all.
Neither picture is a proof device on its own. They make the cases legible; the pivot structure of the reduced system is what settles them.

Forms this is expressed in

The same content in several forms. Each makes something visible that the others leave implicit, so moving between them is part of understanding the topic rather than a presentation choice.

geometric

Two equations as two lines, meeting where both hold

The system $A x = b$ drawn in the space of the unknowns, $R^{n}$ . Each equation $a_{i}^{T} x = b_{i}$ becomes a single flat, a line in two unknowns, a plane in three, a hyperplane in general, consisting of every point satisfying that one equation. The solution set of the whole system is what survives when all $m$ flats are intersected.

The three cases become three pictures. Flats crossing at a single point is the unique solution. Flats that are parallel and distinct never meet, which is inconsistency. Flats that coincide, or meet along a shared line or plane, give infinitely many solutions, and the dimension of what they share is the number of free variables.

This form makes the location of solutions visible, and it is where the feasible region of a linear program lives: replacing each equation by an inequality keeps one side of each flat instead of the flat itself, and the region is the intersection of those sides. What it cannot do is answer existence cheaply, deciding whether $m$ flats in $R^{n}$ share a point generally requires the elimination it was meant to illuminate, and beyond three unknowns there is nothing to draw.

Translates into: geometric

geometric

The same system as a recipe: one A₁ and two A₂ reach b

The same system $A x = b$ drawn in the space of the right-hand side, $R^{m}$ . The columns $A_{1}, \dots, A_{n}$ of $A$ become $n$ arrows, and the question is whether they can be scaled and added to land exactly on $b$ :

x_{1} A_{1} + x_{2} A_{2} + \dots + x_{n} A_{n} = b .

The unknowns are no longer coordinates of a point but weights on arrows. Solving means finding a recipe; the three cases are three answers to whether a recipe exists and whether it is unique.

Columns pointing in independent directions reach a space of dimension equal to their number, and within it every target has exactly one recipe. Columns that are dependent, one a multiple of another, or one a combination of the rest, reach a smaller set, so a target off that set has no recipe at all while a target on it has infinitely many.

This form answers existence directly, because reachability is the question it asks. It is also where the simplex method lives: swapping one column for another changes which set of arrows is available and therefore what can be produced. What it does not show is where the solutions sit as points, which is the row picture's subject. The two are the same system, grouped differently.

Translates into: geometric

geometric

the normals decide whether the lines meet, and how often

the normals decide whether the lines meet, and how often

Three separate coordinate frames side by side, each with its own marked origin and axis stubs labelled $x$ and $y$ ; the offsets between frames are page layout and carry no coordinate meaning. Each frame draws two lines. In the first, $x + 2 y = 5$ and $3 x - y = 1$ have normals pointing in different directions and cross at the single marked point $(1, 2)$ : the system has a unique solution. In the second, $x + 2 y = 5$ and $x + 2 y = 9$ share a normal and differ only in their right-hand sides, so they run parallel and never meet: no solution. In the third, $x + 2 y = 5$ and $2 x + 4 y = 10$ are the same line written twice, drawn as one heavy stroke overlaid by a thinner one: every point of that line solves both equations, so there are infinitely many solutions.

The three things two lines in a plane can do, drawn side by side.

One intersection. $x + 2 y = 5$ and $3 x - y = 1$ have normals $(1, 2)$ and $(3, - 1)$ pointing in different directions, so the lines cross exactly once, at $(1, 2)$ . The system has a unique solution.

No intersection. $x + 2 y = 5$ and $x + 2 y = 9$ have the same normal and different right-hand sides, so they are parallel and distinct. Nothing satisfies both and the system is inconsistent.

Infinitely many intersections. $x + 2 y = 5$ and $2 x + 4 y = 10$ are the same line written twice, because the second equation is the first doubled. Every point of that line solves the system, and the one free variable is the dimension of what the flats share.

Reading the algebra off the geometry is the point: the normals decide whether the lines are parallel, and the right-hand sides then decide whether parallel means disjoint or identical.

Translates into: geometric

geometric

Reachable once, not at all, or in infinitely many ways

Reachable once, not at all, or in infinitely many ways

Three separate coordinate frames side by side, each with its own marked origin and axis stubs labelled "first" and "second" coordinate; the offsets between frames are page layout and carry no coordinate meaning. Each frame draws arrows from its own origin. In the first, the independent columns $A_{1} = (1, 3)$ and $A_{2} = (2, - 1)$ point in different directions and so span the plane; the target $b = (5, 1)$ is drawn in red and is reached by exactly one recipe, $1 A_{1} + 2 A_{2}$ . In the second, the dependent columns $A_{1} = (1, 1)$ and $A_{2} = (2, 2)$ both lie along a single line through the origin, and the target $(1, 4)$ lies off that line, so no combination of the columns reaches it and the system has no solution. In the third, the same dependent columns are shown with the target $(3, 3)$ , which does lie on their common line; it is reachable, but by infinitely many different recipes, since any amount added along one column can be removed along the other.

Whether a target is reachable, in three pictures.

Inside the span. The independent columns $A_{1} = (1, 3)$ and $A_{2} = (2, - 1)$ point in different directions, so together they reach every point of the plane. The target $b = (5, 1)$ is reached by exactly one recipe, $1 A_{1} + 2 A_{2}$ , and there is no other.

Outside the span. Take the dependent pair $A_{1} = (1, 1)$ and $A_{2} = (2, 2)$ . Both lie along one line, so every combination lands on that line and nothing else is reachable. A target off the line, such as $(1, 4)$ , has no recipe at all: the system is inconsistent.

On a dependent span. With the same dependent columns and the target $(3, 3)$ , which does lie on the line, the recipe is no longer unique: $3 A_{1} + 0 A_{2}$ and $1 A_{1} + 1 A_{2}$ both reach it, and so do infinitely many others.

Existence and uniqueness separate here in a way the row picture does not show: whether the columns span enough decides existence, and whether they are independent decides uniqueness.

Translates into: geometric

Worked material

Example

One system, both pictures

Take

\begin{aligned} x + 2 y & = 5 \\ 3 x - y & = 1 \end{aligned}

Row picture. Two lines in the plane: $x + 2 y = 5$ and $3 x - y = 1$ . They are not parallel, their normals $(1, 2)$ and $(3, - 1)$ point in different directions, so they cross at exactly one point. Solving gives $(1, 2)$ .

Column picture. The columns are $A_{1} = (1, 3)$ and $A_{2} = (2, - 1)$ , and the question is which weights reach $b = (5, 1)$ :

x (\begin{matrix} 1 \\ 3 \end{matrix}) + y (\begin{matrix} 2 \\ - 1 \end{matrix}) = (\begin{matrix} 5 \\ 1 \end{matrix}) .

With $x = 1$ and $y = 2$ : $(1, 3) + (4, - 2) = (5, 1)$ . The same answer, arrived at by mixing rather than by intersecting.

What each made visible. The row picture showed where the solution sits. A point in the plane of the unknowns. The column picture showed why one exists: the two column arrows point in independent directions, so together they reach everything in $R^{2}$ , and $b$ was never in doubt.

A one-word change. Make the second column $(2, 6)$ instead. Now both columns lie along the direction $(1, 3)$ , they can only reach that line, and whether a solution exists depends entirely on whether $b$ happens to lie on it. The column picture settles that instantly; the row picture needs elimination to discover the lines are parallel.

Non-example

Things the two pictures do not establish

Treating them as a matter of taste. They answer different questions with different effort. Choosing by preference rather than by question is what makes half the subject feel obscure.

Expecting them in the same space. The row picture lives in $R^{n}$ , the column picture in $R^{m}$ . For a $3 \times 2$ system one diagram is in the plane and the other in three dimensions.

Reading parallel rows as dependent columns. In the square case the two coincide; in general the link runs through rank, not through a visual analogy between the drawings.

Using a picture as a proof. They make the case legible and memorable. The pivot structure of the reduced system is what decides it.

Assuming a disagreement is a discovery. Both describe the same solution set. If they seem to disagree, one of them has been drawn wrong.

Drawing the column picture for large $n$ . With twenty columns there is nothing to draw. The reading still works as a way of thinking, which vectors can be reached, long after the diagram has stopped being possible.

Contrast

Which picture answers which question

	Row picture	Column picture
Each equation is	A hyperplane	One column's contribution
The solution set is	Where the flats intersect	The weights that reach $b$
Lives in	$R^{n}$ , the unknowns	$R^{m}$ , the right-hand side
Answers easily	Where is the solution?	Does one exist?
Reappears as	The feasible region	Bases, spans, pivots
Number of objects drawn	One per equation	One per variable

The asymmetry to remember. Existence is a column question. Location is a row question. Asking the wrong picture for the wrong one is not an error, merely slow, and sometimes very slow, as when a five-second observation about a multiple replaces a full elimination.

Where each is met again. A graphical solution of a linear program draws the row picture and slides a contour across it. The simplex method works the column picture, swapping one column for another and asking what the new mixture reaches. Both appear in this subject within a few lessons of each other.

The same solution set, always. The two readings group the same equation differently, so the solution set is identical by construction rather than by coincidence.

Common errors

Common misconception

The row picture and the column picture are two different ways of drawing a system, so a learner may simply use whichever they find easier.

Related units

Requires

Solving and Characterising Linear Systems

Connected

The Feasible Region of a Linear Program (part of)
Half-Spaces and Hyperplanes (related)

Learn this topic

The Row and Column Pictures

Used in

Foundations for Operational Research

Sources

Introduction to Linear Algebra (2016)
Introduction to Linear Optimization (1997)

The Row and Column Pictures

Assumptions and scope

Forms this is expressed in

Worked material

One system, both pictures

Things the two pictures do not establish

Which picture answers which question

Common errors

Related units

Requires

Connected

Learn this topic

Used in

Sources

Settings

Focus timer

Today

Last seven days

Where the time went

Recent sessions

Settings

Notifications

Your history