SVD Calculator

Key formulas for the 2×2 SVD

In general, the singular values of $A$ are the square roots of the eigenvalues of $A^{T}A$. For a 2×2 real matrix, this can be written out explicitly. Let

Formula: A^T A = b_11 b_12 b_12 b_22.

$A^{T}A=\left[\begin{array}{cc}{b}_{11}& b_{12}\\ b_{12}& b_{22}\end{array}\right].$

Then the eigenvalues ${\lambda }_1$ and ${\lambda }_2$ come from the characteristic polynomial of $A^{T}A$. The basic relationship between eigenvalues and singular values is

Here $\lambda$ is an eigenvalue of $A^{T}A$, and the corresponding singular value is $\sigma =\sqrt{\lambda }$. For a 2×2 matrix, you always get up to two singular values ${\sigma }_1\ge {\sigma }_2\ge 0$.

A convenient closed form for the 2×2 case is

Formula: σ_1,2 = sqrt((tr ⁡ (A^T A) ± sqrt(tr⁡(A^TA)^2 - 4 det ⁡ (A^T A))) / 2)

$${\sigma }_{1,2}=\sqrt{\frac{\mathrm{tr}\left(A^{T}A\right)\pm \sqrt{{\mathrm{tr}\left(A^{T}A\right)}^2-4\det \left(A^{T}A\right)}}{2}}$$

Once the calculator has the singular values, it computes the right singular vectors, which are the columns of $V$, from the eigenvectors of $A^{T}A$. Finally, it finds the left singular vectors, the columns of $U$, from

Formula: U = A V Σ^−1,

$U=AV{\Sigma }^{-1},$

with appropriate handling of any zero singular values. That last detail matters for rank-deficient or nearly rank-deficient matrices, where one direction is stretched by almost nothing at all.

Geometric interpretation for 2×2 matrices

Because this tool focuses on 2×2 matrices, you can picture everything in the familiar two-dimensional plane. That makes SVD unusually intuitive here. A matrix does not merely shuffle numbers; it sends the unit circle to a new shape in the output plane.

• Start with the unit circle in the input space, meaning all points with distance 1 from the origin.
• Apply the matrix $A$. The image of that circle is generally an ellipse in the output space.
• The lengths of the semi-axes of that ellipse are the singular values ${\sigma }_1$ and ${\sigma }_2$.
• The directions of those axes are given by the columns of $U$ for the output and the columns of $V$ for the input.

In this view, SVD tells you exactly how your matrix stretches and rotates the plane. Large singular values correspond to directions where vectors are strongly stretched. If the smallest singular value is close to zero, there is a direction that is almost collapsed, meaning the transformation nearly loses dimensionality. That geometric picture is one of the main reasons SVD is so important: it reveals both rank and conditioning in a single glance.

How to use this 2×2 SVD calculator

1. Enter the four entries of your 2×2 matrix into the fields labeled $a_{11}$, $a_{12}$, $a_{21}$, and $a_{22}$.
2. Click the Compute SVD button.
3. Review the results: the calculator shows $U$, $\Sigma$, and $V^T$. It also gives you a quick rank estimate and a simple conditioning indicator based on the ratio of the singular values.

This is especially helpful when the arithmetic is messy. By automating the eigenvalue and normalization steps, the calculator lets you spend your attention on interpretation instead of bookkeeping.

Interpreting the results

After you compute the SVD, focus on three main pieces of information. The most useful habit is to read the decomposition as a story: first identify the important directions, then read how strongly each one is stretched, and finally note the output orientation.

1. Singular values in $\Sigma$

The diagonal entries ${\sigma }_1$ and ${\sigma }_2$ of $\Sigma$ tell you how much the matrix stretches vectors along certain orthogonal directions.

• If both singular values are far from zero, the matrix is well-conditioned and does not collapse any direction strongly.
• If ${\sigma }_2$ is very small compared with ${\sigma }_1$, the transformation is close to rank 1 and maps most vectors into something close to a line.
• If one singular value is exactly zero, the matrix is rank-deficient and not invertible.

2. Left singular vectors in $U$

The columns of $U$ are unit vectors in the output space. Each column corresponds to a principal direction of the ellipse produced by transforming the unit circle. The first column aligns with the direction of maximum stretching, and the second column is perpendicular to it.

3. Right singular vectors in $V$

The columns of $V$ are unit vectors in the input space. They represent the directions that, when you apply $A$, are scaled by the corresponding singular values and rotated into the directions given by $U$.

Together, these pieces let you describe the action of $A$ in a compact and memorable sequence: rotate by $V^T$, stretch by $\Sigma$, then rotate by $U$.

Worked example: SVD of a 2×2 matrix

Consider the matrix

Formula: A = [3 1 0 2].

$A=\left[\begin{array}{cc}3& 1\\ 0& 2\end{array}\right].$

Below is the outline of the steps the calculator conceptually follows.

Step 1: Form $A^{T}A$

Formula: A^T = [3 0 1 2], A^T A = [3 0 1 2] [3 1 0 2] = [9 3 3 5].

$A^T=\left[\begin{array}{cc}3& 0\\ 1& 2\end{array}\right],A^{T}A=\left[\begin{array}{cc}3& 0\\ 1& 2\end{array}\right]\left[\begin{array}{cc}3& 1\\ 0& 2\end{array}\right]=\left[\begin{array}{cc}9& 3\\ 3& 5\end{array}\right].$

Step 2: Find eigenvalues of $A^{T}A$

The characteristic polynomial of $A^{T}A$ is

Formula: det ⁡ (A^T ⁢ A ⁢ - ⁢ λ ⁢ I) = | 9 - λ 3 3 5 - λ | = (9 - λ)(5 - λ) - 9.

$\det (A^{T}A-\lambda I)=\left|\begin{array}{cc}9-\lambda & 3\\ 3& 5-\lambda \end{array}\right|=(9-\lambda )(5-\lambda )-9.$

Simplifying gives

Formula: λ^2 − 14 λ + 36 = 0.

${\lambda }^2-14\lambda +36=0.$

Solving this quadratic equation yields two eigenvalues ${\lambda }_1$ and ${\lambda }_2$. Numerically, these are approximately

Formula: λ_1 ≈ 11.944, λ_2 ≈ 2.056.

${\lambda }_1\approx 11.944,{\lambda }_2\approx 2.056.$

Step 3: Compute singular values

Using $\sigma =\sqrt{\lambda }$, the singular values are approximately

Formula: σ_1 = sqrt(11.944) ≈ 3.456, σ_2 = sqrt(2.056) ≈ 1.434.

${\sigma }_1=\sqrt{11.944}\approx 3.456,{\sigma }_2=\sqrt{2.056}\approx 1.434.$

Step 4: Find eigenvectors (right singular vectors)

For each eigenvalue, solve $(A^{T}A-\lambda I)v=0$ to find a nonzero eigenvector $v$. Normalize each eigenvector to have length 1. These normalized eigenvectors form the columns of $V$. The calculator performs this step numerically so you do not have to.

Step 5: Compute left singular vectors

For each singular value ${\sigma }_i$ and right singular vector $v_i$, compute

Formula: u_i = 1 / σ_i A v_i.

$u_i=\frac{1}{{\sigma }_i}A{v}_i.$

These vectors are then normalized as needed and collected as the columns of $U$. The final result displayed by the calculator is the factorization $A=U\Sigma V^T$, with approximate decimal entries for ease of interpretation.

Comparison: SVD vs. other matrix decompositions

The SVD is related to other common decompositions but not identical. The comparison below summarizes what each factorization is best at revealing for small real matrices.

Assumptions and limitations of this calculator

• Matrix size: The tool is designed for 2×2 matrices only. It does not accept larger or rectangular matrices.
• Real entries only: Inputs are assumed to be real numbers. Complex matrices and complex-valued SVD are not supported.
• Numerical rounding: Results are shown with finite decimal precision. Very large or very small values may be subject to rounding errors.
• Near-singular matrices: When singular values are extremely small, distinguishing between zero and almost zero depends on numerical tolerance. Interpret such results with care.
• Educational focus: The calculator is intended for learning, checking homework, and building intuition. It is not a substitute for high-precision numerical libraries in critical engineering or scientific applications.
• No symbolic algebra: The tool does not produce symbolic expressions; it evaluates numeric SVDs based on the numbers you enter.

Those limitations are normal for a compact browser calculator. They do not reduce the value of the page for study; they simply clarify the scope. The 2×2 setting is often the best place to learn the meaning of singular vectors and singular values before scaling up to larger software tools.

Tips for using the SVD in practice

• Use the ratio $\frac{{\sigma }_1}{{\sigma }_2}$ as a simple indicator of conditioning for 2×2 matrices. A very large ratio suggests numerical sensitivity.
• Experiment with different inputs to see how changing one entry of the matrix affects the singular values and singular vectors.
• Compare SVD-based insights with the determinant and trace of your matrix to see different perspectives on the same linear transformation.

As a quick rule of thumb, large separation between the singular values means one direction dominates, while nearly equal singular values mean the transformation behaves more like a uniform scaling plus rotation.

Applications of SVD

Even though this calculator works with small 2×2 matrices, the same ideas extend to much larger systems. SVD is central in data analysis and PCA, least squares, regression, image compression, signal processing, and many numerical optimization methods. In each of those settings, the decomposition helps separate dominant structure from weaker directions.

• Data analysis and PCA: SVD underlies principal component analysis, which finds directions of maximum variance in high-dimensional data.
• Least squares and regression: The pseudoinverse built from SVD is a robust way to solve overdetermined systems and linear regression problems.
• Image compression: Large image matrices can be approximated by keeping only the largest singular values, reducing storage while keeping key features.
• Signal processing: SVD helps separate signal from noise by focusing on dominant singular values.

Studying the 2×2 case with this tool is a good way to build intuition before moving on to larger matrices. When you can see a unit circle become an ellipse, the abstract notation $A=U\Sigma V^T$ becomes much easier to remember and use correctly.