Row Picture Vs Column Picture

Consider the following system of 2 equations:

Conventionally, we interpret this system as representing two lines in a two-dimensional plane. If the lines intersect, the point of intersection corresponds to a unique solution. If the lines do not intersect—because they are parallel—the system may have no solution or infinitely many solutions, depending on whether the lines are distinct or coincide.

We may also express this system using matrix notation:

However, this notation alone does not fundamentally change our geometric understanding. To gain a different perspective, we can rewrite the system as follows:

In this form, the variables x and y appear as scalars multiplying two vectors. These vectors arise directly from the matrix representation: they are the first and second columns of the matrix A. Let us denote them by v₁ and v₂​.

Mathematically, this formulation is equivalent to the original system. Conceptually, however, the question changes. Instead of asking where two lines intersect, we now ask:

This provides a different but equivalent interpretation of the same problem.

From this viewpoint, we can again ask under what conditions the system has a unique solution, infinitely many solutions, or no solution. The answer depends on the relationship between v₁ and v₂, and also b.

• If v₁ and v₂​​​ are not parallel—that is, if each has at least some component not aligned with the other—then they span the entire two-dimensional plane. In this case, for any vector b, there exists a unique solution.

• If v₁ and v₂​ are parallel, they span only a one-dimensional subspace. we call such matrices, singular. In such cases, the system then has:

  • infinitely many solutions if b lies along the same direction, or

  • no solution if b does not.

For large systems of n equations with n unknowns, which are common in engineering applications, the geometric interpretation based on line intersections is no longer informative.

However, this vector-based viewpoint raises additional questions that are worth clarifying. In particular, in a system of equations, we can freely change the order of the equations without affecting the solution. Yet, in the column-vector interpretation, the columns of the matrix appear to play a fixed structural role. How can these two facts be consistent?

The answer lies in understanding the effect of row operations. Reordering equations corresponds to applying a linear transformation to the rows of the system. Algebraically, this is equivalent to multiplying both sides of the equation by a permutation matrix, which is always invertible. Since invertible linear transformations preserve solution sets, the solution of the system remains unchanged.

Importantly, this operation changes only how the equations are represented; it does not alter the underlying linear relationship between the variables. In the column-vector interpretation, the column vectors still represent the contributions of each variable to the output. What changes is merely how the resulting vectors are expressed numerically.

This reasoning extends beyond simple row exchanges. Any linear combination of rows—such as adding or subtracting equations—can be represented as multiplication by an invertible matrix. These operations form the basis of Gaussian elimination.

For example, adding half of the first equation to the second corresponds to multiplying both sides by a matrix as below

Although this transformation changes the appearance of the coefficient matrix, it does not change the solution. The abstract vectors represented by the columns remain the same; only their coordinate descriptions are altered, but all linear dependence relations among them are preserved.

The vector-based perspective provides a practical framework for further analysis and reduces the problem to properties of the matrix A (rank, span, singularity, etc.). By examining the coefficient matrix A, we can determine whether a unique solution is possible. Also, when A is singular, this approach allows us to investigate for which vector b the system admits solutions.