5 HYPERPLANES AND HALF-SPACES

5.1. Hyperplanes

hyperplane is a set described by a single scalar product equality. Precisely, a hyperplane in \mathbb{R}^n is a set of the form

    \begin{align*} \mathbf{H} = \{x: a^Tx = b\}, \end{align*}

where a \in \mathbb{R}^n, a \neq 0, and b \in \mathbb{R} are given. When b=0, the hyperplane is simply the set of points that are orthogonal to a; when b \neq 0, the hyperplane is a translation, along direction a, of that set.

If x_0 \in {\bf H}, then for any other element x \in {\bf H}, we have

    \begin{align*} b = a^Tx_0 = a^Tx. \end{align*}

Hence, the hyperplane can be characterized as the set of vectors x such that x - x_0 is orthogonal to a:

    \begin{align*} \mathbf{H} = \{x: a^T(x-x_0) = 0\}. \end{align*}

Hyperplanes are affine sets, of dimension n-1 (see the proof here). Thus, they generalize the usual notion of a plane in \mathbb{R}^3. Hyperplanes are very useful because they allows to separate the whole space in two regions. The notion of half-space formalizes this.

Example 1: A hyperplane in \mathbb{R}^3.
Consider an affine set of dimension 2 in \mathbb{R}^3, which we describe as the set of points x in \mathbb{R}^3 such that there exists two parameters \lambda_1, \lambda_2 such that

    \[ x = \begin{pmatrix} 3\lambda_1-4\lambda_2 + 4 \\ \lambda_1 \\ \lambda_2 \end{pmatrix} = \begin{pmatrix} 4 \\ 0 \\ 0 \end{pmatrix} + \lambda_1 \begin{pmatrix} 3 \\ 1 \\ 0 \end{pmatrix} + \lambda_2 \begin{pmatrix} -4 \\ 0 \\ 1 \end{pmatrix}. \]

The set \mathbf{H} can be represented as a translation of a linear subspace: \mathbf{H} = x_0 + \mathbf{L}, with

    \[ x_0 := \begin{pmatrix} 4 \\ 0 \\ 0 \end{pmatrix}, \]

and \mathbf{L} the span of the two independent vectors

    \[ u := \begin{pmatrix} 3 \\ 1 \\ 0 \end{pmatrix}, \quad v := \begin{pmatrix} -4 \\ 0 \\ 1 \end{pmatrix}. \]

Thus, the set \mathbf{H} is of dimension 2 in \mathbb{R}^3, hence it is a hyperplane. In \mathbb{R}^3, hyperplanes are ordinary planes. We can find a representation of the hyperplane in the standard form

    \[ \mathbf{H} = \left\{ x \,:\, a^T(x-x_0) = 0 \right\}. \]

We simply find a that is orthogonal to both u and v. That is, we solve the equations

    \[ 0 = a^T u = 3a_1 + a_2, \quad 0 = a^T v = -4a_1 + a_3. \]

The above leads to a = (a_1, -3a_1, 4a_1). Choosing for example a_1 = 1 leads to a=(1,-3,4).
The hyperplane \mathbf{H} can be expressed as x_0 + \mathbf{span}(u, v), where x_0 is a particular element, and u, v are two independent vectors. The set \mathbf{H} is represented in light blue; it is a translation of the corresponding span \mathbf{L} = \mathbf{span}(u, v). Any point x in \mathbf{H} is such that x - x_0 belongs to \mathbf{L}. Thus, we can represent the hyperplane as the set of points such that x - x_0 is orthogonal to a, where a is any vector orthogonal to both u, v.

5.2. Projection on a hyperplane

Consider the hyperplane {\bf H} = \{x: a^Tx = b\}, and assume without loss of generality that a is normalized (||a||_2 =1). We can represent {\bf H} as the set of points x such that x- x_0 is orthogonal to a, where x_0 is any vector in {\bf H}, that is, such that a^Tx_0 = b. One such vector is x_{proj}:= ba.

By construction, x_{proj} is the projection of 0 on {\bf H}. That is, it is the point on {\bf H} closest to the origin, as it solves the projection problem

    \begin{align*} \min\limits_x ||x||_2: x \in \mathbf{H} \end{align*}

Indeed, for any  x \in {\bf H}, using the Cauchy-Schwartz inequality:

    \begin{align*} ||x_0||_2 = |b| = |a^Tx| \leq ||a||_2 \cdot ||x||_2 = ||x||_2, \end{align*}

and the minimum length |b| is attained with x_{proj} = ba.

5.3. Geometry of hyperplanes

Geometrically, a hyperplane {\bf H} = \{x: a^Tx = b\}, with ||a||_2 = 1, is a translation of the set of vectors orthogonal to a. The direction of the translation is determined by a, and the amount by b.
Precisely, |b| is the length of the closest point x_0 on {\bf H} from the origin, and the sign of b determines if {\bf H} is away from the origin along the direction a or -a. As we increase the magnitude of b, the hyperplane is shifting further away along \pm a, depending on the sign of b.
In a 3D space, a hyperplane corresponds to a plane. In the image on the left, the scalar b is positive, as x_0 and a point to the same direction.

5.4. Half-spaces

A half-space is a subset of \mathbb{R}^n defined by a single inequality involving a scalar product. Precisely, a half-space in \mathbb{R}^n is a set of the form

    \begin{align*} \mathbf{H} = \{x: a^Tx \ge b\}, \end{align*}

where a \in \mathbb{R}^n, a \neq 0, and b \in \mathbb{R} are given.

Geometrically, the half-space above is the set of points such that {a^T(x-x_0) \ge 0, that is, the angle between x - x_0 and a is acute (in [-90^{\circ}; +90^{\circ}]). Here x_0 is the point closest to the origin on the hyperplane defined by the equality a^Tx = b. (When a is normalized, as in the picture, x_0 = ba.)

The half-space \{x: a^Tx \ge b\}, is the set of points such that x-x_0 forms an acute angle with a, where x_0 is the projection of the origin on the boundary of the half-space.
The image on the left is a visualization of half-spaces in a 3D context.

License

Icon for the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License

Linear Algebra and Applications Copyright © 2023 by VinUiversity is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, except where otherwise noted.

Share This Book