Lp space

In mathematics, the L^p spaces are function spaces defined using a natural generalization of the p-norm for finite-dimensional vector spaces. They are sometimes called Lebesgue spaces, named after Henri Lebesgue (Dunford & Schwartz 1958, III.3), although according to the Bourbaki group (Bourbaki 1987) they were first introduced by Frigyes Riesz (Riesz 1910).

L^p spaces form an important class of Banach spaces in functional analysis, and of topological vector spaces. Because of their key role in the mathematical analysis of measure and probability spaces, Lebesgue spaces are used also in the theoretical discussion of problems in physics, statistics, economics, finance, engineering, and other disciplines.

In statistics, measures of central tendency and statistical dispersion, such as the mean, median, and standard deviation, can be defined in terms of ${\displaystyle L^{p}}$ metrics, and measures of central tendency can be characterized as solutions to variational problems.

In penalized regression, "L1 penalty" and "L2 penalty" refer to penalizing either the ${\displaystyle L^{1}}$ norm of a solution's vector of parameter values (i.e. the sum of its absolute values), or its squared ${\displaystyle L^{2}}$ norm (its Euclidean length). Techniques which use an L1 penalty, like LASSO, encourage sparse solutions (where the many parameters are zero).^[1] Elastic net regularization uses a penalty term that is a combination of the ${\displaystyle L^{1}}$ norm and the squared ${\displaystyle L^{2}}$ norm of the parameter vector.

The Fourier transform for the real line (or, for periodic functions, see Fourier series), maps ${\displaystyle L^{p}(\mathbb {R} )}$ to ${\displaystyle L^{q}(\mathbb {R} )}$ (or ${\displaystyle L^{p}(\mathbf {T} )}$ to ${\displaystyle \ell ^{q}}$) respectively, where ${\displaystyle 1\leq p\leq 2}$ and ${\displaystyle {\tfrac {1}{p}}+{\tfrac {1}{q}}=1.}$ This is a consequence of the Riesz–Thorin interpolation theorem, and is made precise with the Hausdorff–Young inequality.

By contrast, if ${\displaystyle p>2,}$ the Fourier transform does not map into ${\displaystyle L^{q}.}$

Hilbert spaces are central to many applications, from quantum mechanics to stochastic calculus. The spaces ${\displaystyle L^{2}}$ and ${\displaystyle \ell ^{2}}$ are both Hilbert spaces. In fact, by choosing a Hilbert basis ${\displaystyle E,}$ i.e., a maximal orthonormal subset of ${\displaystyle L^{2}}$ or any Hilbert space, one sees that every Hilbert space is isometrically isomorphic to ${\displaystyle \ell ^{2}(E)}$ (same ${\displaystyle E}$ as above), i.e., a Hilbert space of type ${\displaystyle \ell ^{2}.}$

The Euclidean length of a vector ${\displaystyle x=(x_{1},x_{2},\dots ,x_{n})}$ in the ${\displaystyle n}$-dimensional real vector space ${\displaystyle \mathbb {R} ^{n}}$ is given by the Euclidean norm: $${\displaystyle \|x\|_{2}=\left({x_{1}}^{2}+{x_{2}}^{2}+\dotsb +{x_{n}}^{2}\right)^{1/2}.}$$

The Euclidean distance between two points ${\displaystyle x}$ and ${\displaystyle y}$ is the length ${\displaystyle \|x-y\|_{2}}$ of the straight line between the two points. In many situations, the Euclidean distance is appropriate for capturing the actual distances in a given space. In contrast, consider taxi drivers in a grid street plan who should measure distance not in terms of the length of the straight line to their destination, but in terms of the rectilinear distance, which takes into account that streets are either orthogonal or parallel to each other. The class of ${\displaystyle p}$-norms generalizes these two examples and has an abundance of applications in many parts of mathematics, physics, and computer science.

For a real number ${\displaystyle p\geq 1,}$ the ${\displaystyle p}$-norm or ${\displaystyle L^{p}}$-norm of ${\displaystyle x}$ is defined by $${\displaystyle \|x\|_{p}=\left(|x_{1}|^{p}+|x_{2}|^{p}+\dotsb +|x_{n}|^{p}\right)^{1/p}.}$$ The absolute value bars can be dropped when ${\displaystyle p}$ is a rational number with an even numerator in its reduced form, and ${\displaystyle x}$ is drawn from the set of real numbers, or one of its subsets.

The Euclidean norm from above falls into this class and is the ${\displaystyle 2}$-norm, and the ${\displaystyle 1}$-norm is the norm that corresponds to the rectilinear distance.

The ${\displaystyle L^{\infty }}$-norm or maximum norm (or uniform norm) is the limit of the ${\displaystyle L^{p}}$-norms for ${\displaystyle p\to \infty .}$ It turns out that this limit is equivalent to the following definition: $${\displaystyle \|x\|_{\infty }=\max \left\{|x_{1}|,|x_{2}|,\dotsc ,|x_{n}|\right\}}$$

See L-infinity.

For all ${\displaystyle p\geq 1,}$ the ${\displaystyle p}$-norms and maximum norm as defined above indeed satisfy the properties of a "length function" (or norm), which are that:

• only the zero vector has zero length,
• the length of the vector is positive homogeneous with respect to multiplication by a scalar (positive homogeneity), and
• the length of the sum of two vectors is no larger than the sum of lengths of the vectors (triangle inequality).

Abstractly speaking, this means that ${\displaystyle \mathbb {R} ^{n}}$ together with the ${\displaystyle p}$-norm is a normed vector space. Moreover, it turns out that this space is complete, thus making it a Banach space. This Banach space is the ${\displaystyle L^{p}}$-space over ${\displaystyle \{1,2,\ldots ,n\}.}$

The grid distance or rectilinear distance (sometimes called the "Manhattan distance") between two points is never shorter than the length of the line segment between them (the Euclidean or "as the crow flies" distance). Formally, this means that the Euclidean norm of any vector is bounded by its 1-norm: $${\displaystyle \|x\|_{2}\leq \|x\|_{1}.}$$

This fact generalizes to ${\displaystyle p}$-norms in that the ${\displaystyle p}$-norm ${\displaystyle \|x\|_{p}}$ of any given vector ${\displaystyle x}$ does not grow with ${\displaystyle p}$:

${\displaystyle \|x\|_{p+a}\leq \|x\|_{p}}$ for any vector ${\displaystyle x}$ and real numbers ${\displaystyle p\geq 1}$ and ${\displaystyle a\geq 0.}$ (In fact this remains true for ${\displaystyle 0<p<1}$ and ${\displaystyle a\geq 0}$ .)

For the opposite direction, the following relation between the ${\displaystyle 1}$-norm and the ${\displaystyle 2}$-norm is known: $${\displaystyle \|x\|_{1}\leq {\sqrt {n}}\|x\|_{2}~.}$$

This inequality depends on the dimension ${\displaystyle n}$ of the underlying vector space and follows directly from the Cauchy–Schwarz inequality.

In general, for vectors in ${\displaystyle \mathbb {C} ^{n}}$ where ${\displaystyle 0<r<p:}$ $${\displaystyle \|x\|_{p}\leq \|x\|_{r}\leq n^{{\frac {1}{r}}-{\frac {1}{p}}}\|x\|_{p}~.}$$

This is a consequence of Hölder's inequality.

In ${\displaystyle \mathbb {R} ^{n}}$ for ${\displaystyle n>1,}$ the formula $${\displaystyle \|x\|_{p}=\left(|x_{1}|^{p}+|x_{2}|^{p}+\cdots +|x_{n}|^{p}\right)^{1/p}}$$ defines an absolutely homogeneous function for ${\displaystyle 0<p<1;}$ however, the resulting function does not define a norm, because it is not subadditive. On the other hand, the formula $${\displaystyle |x_{1}|^{p}+|x_{2}|^{p}+\dotsb +|x_{n}|^{p}}$$ defines a subadditive function at the cost of losing absolute homogeneity. It does define an F-norm, though, which is homogeneous of degree ${\displaystyle p.}$

Hence, the function $${\displaystyle d_{p}(x,y)=\sum _{i=1}^{n}|x_{i}-y_{i}|^{p}}$$ defines a metric. The metric space ${\displaystyle (\mathbb {R} ^{n},d_{p})}$ is denoted by ${\displaystyle \ell _{n}^{p}.}$

Although the ${\displaystyle p}$-unit ball ${\displaystyle B_{n}^{p}}$ around the origin in this metric is "concave", the topology defined on ${\displaystyle \mathbb {R} ^{n}}$ by the metric ${\displaystyle B_{p}}$ is the usual vector space topology of ${\displaystyle \mathbb {R} ^{n},}$ hence ${\displaystyle \ell _{n}^{p}}$ is a locally convex topological vector space. Beyond this qualitative statement, a quantitative way to measure the lack of convexity of ${\displaystyle \ell _{n}^{p}}$ is to denote by ${\displaystyle C_{p}(n)}$ the smallest constant ${\displaystyle C}$ such that the scalar multiple ${\displaystyle C\,B_{n}^{p}}$ of the ${\displaystyle p}$-unit ball contains the convex hull of ${\displaystyle B_{n}^{p},}$ which is equal to ${\displaystyle B_{n}^{1}.}$ The fact that for fixed ${\displaystyle p<1}$ we have $${\displaystyle C_{p}(n)=n^{{\tfrac {1}{p}}-1}\to \infty ,\quad {\text{as }}n\to \infty }$$ shows that the infinite-dimensional sequence space ${\displaystyle \ell ^{p}}$ defined below, is no longer locally convex.^{[citation needed]}

There is one ${\displaystyle \ell _{0}}$ norm and another function called the ${\displaystyle \ell _{0}}$ "norm" (with quotation marks).

The mathematical definition of the ${\displaystyle \ell _{0}}$ norm was established by Banach's Theory of Linear Operations. The space of sequences has a complete metric topology provided by the F-norm $${\displaystyle (x_{n})\mapsto \sum _{n}2^{-n}{\frac {|x_{n}|}{1+|x_{n}|}},}$$ which is discussed by Stefan Rolewicz in Metric Linear Spaces.^[2] The ${\displaystyle \ell _{0}}$-normed space is studied in functional analysis, probability theory, and harmonic analysis.

Another function was called the ${\displaystyle \ell _{0}}$ "norm" by David Donoho—whose quotation marks warn that this function is not a proper norm—is the number of non-zero entries of the vector ${\displaystyle x.}$^{[citation needed]} Many authors abuse terminology by omitting the quotation marks. Defining ${\displaystyle 0^{0}=0,}$ the zero "norm" of ${\displaystyle x}$ is equal to $${\displaystyle |x_{1}|^{0}+|x_{2}|^{0}+\cdots +|x_{n}|^{0}.}$$

This is not a norm because it is not homogeneous. For example, scaling the vector ${\displaystyle x}$ by a positive constant does not change the "norm". Despite these defects as a mathematical norm, the non-zero counting "norm" has uses in scientific computing, information theory, and statistics–notably in compressed sensing in signal processing and computational harmonic analysis. Despite not being a norm, the associated metric, known as Hamming distance, is a valid distance, since homogeneity is not required for distances.

The ${\displaystyle p}$-norm can be extended to vectors that have an infinite number of components (sequences), which yields the space ${\displaystyle \ell ^{p}.}$ This contains as special cases:

• ${\displaystyle \ell ^{1},}$ the space of sequences whose series are absolutely convergent,
• ${\displaystyle \ell ^{2},}$ the space of square-summable sequences, which is a Hilbert space, and
• ${\displaystyle \ell ^{\infty },}$ the space of bounded sequences.

The space of sequences has a natural vector space structure by applying addition and scalar multiplication coordinate by coordinate. Explicitly, the vector sum and the scalar action for infinite sequences of real (or complex) numbers are given by: $${\displaystyle {\begin{aligned}&(x_{1},x_{2},\ldots ,x_{n},x_{n+1},\ldots )+(y_{1},y_{2},\ldots ,y_{n},y_{n+1},\ldots )\\={}&(x_{1}+y_{1},x_{2}+y_{2},\ldots ,x_{n}+y_{n},x_{n+1}+y_{n+1},\ldots ),\\[6pt]&\lambda \cdot \left(x_{1},x_{2},\ldots ,x_{n},x_{n+1},\ldots \right)\\={}&(\lambda x_{1},\lambda x_{2},\ldots ,\lambda x_{n},\lambda x_{n+1},\ldots ).\end{aligned}}}$$

Define the ${\displaystyle p}$-norm: $${\displaystyle \|x\|_{p}=\left(|x_{1}|^{p}+|x_{2}|^{p}+\cdots +|x_{n}|^{p}+|x_{n+1}|^{p}+\cdots \right)^{1/p}}$$

Here, a complication arises, namely that the series on the right is not always convergent, so for example, the sequence made up of only ones, ${\displaystyle (1,1,1,\ldots ),}$ will have an infinite ${\displaystyle p}$-norm for ${\displaystyle 1\leq p<\infty .}$ The space ${\displaystyle \ell ^{p}}$ is then defined as the set of all infinite sequences of real (or complex) numbers such that the ${\displaystyle p}$-norm is finite.

One can check that as ${\displaystyle p}$ increases, the set ${\displaystyle \ell ^{p}}$ grows larger. For example, the sequence $${\displaystyle \left(1,{\frac {1}{2}},\ldots ,{\frac {1}{n}},{\frac {1}{n+1}},\ldots \right)}$$ is not in ${\displaystyle \ell ^{1},}$ but it is in ${\displaystyle \ell ^{p}}$ for ${\displaystyle p>1,}$ as the series $${\displaystyle 1^{p}+{\frac {1}{2^{p}}}+\cdots +{\frac {1}{n^{p}}}+{\frac {1}{(n+1)^{p}}}+\cdots ,}$$ diverges for ${\displaystyle p=1}$ (the harmonic series), but is convergent for ${\displaystyle p>1.}$

One also defines the ${\displaystyle \infty }$-norm using the supremum: $${\displaystyle \|x\|_{\infty }=\sup(|x_{1}|,|x_{2}|,\dotsc ,|x_{n}|,|x_{n+1}|,\ldots )}$$ and the corresponding space ${\displaystyle \ell ^{\infty }}$ of all bounded sequences. It turns out that^[3] $${\displaystyle \|x\|_{\infty }=\lim _{p\to \infty }\|x\|_{p}}$$ if the right-hand side is finite, or the left-hand side is infinite. Thus, we will consider ${\displaystyle \ell ^{p}}$ spaces for ${\displaystyle 1\leq p\leq \infty .}$

The ${\displaystyle p}$-norm thus defined on ${\displaystyle \ell ^{p}}$ is indeed a norm, and ${\displaystyle \ell ^{p}}$ together with this norm is a Banach space. The fully general ${\displaystyle L^{p}}$ space is obtained—as seen below—by considering vectors, not only with finitely or countably-infinitely many components, but with "arbitrarily many components"; in other words, functions. An integral instead of a sum is used to define the ${\displaystyle p}$-norm.

In complete analogy to the preceding definition one can define the space ${\displaystyle \ell ^{p}(I)}$ over a general index set ${\displaystyle I}$ (and ${\displaystyle 1\leq p<\infty }$) as $${\displaystyle \ell ^{p}(I)=\left\{(x_{i})_{i\in I}\in \mathbb {K} ^{I}:\sum _{i\in I}|x_{i}|^{p}<+\infty \right\},}$$ where convergence on the right means that only countably many summands are nonzero (see also Unconditional convergence). With the norm $${\displaystyle \|x\|_{p}=\left(\sum _{i\in I}|x_{i}|^{p}\right)^{1/p}}$$ the space ${\displaystyle \ell ^{p}(I)}$ becomes a Banach space. In the case where ${\displaystyle I}$ is finite with ${\displaystyle n}$ elements, this construction yields ${\displaystyle \mathbb {R} ^{n}}$ with the ${\displaystyle p}$-norm defined above. If ${\displaystyle I}$ is countably infinite, this is exactly the sequence space ${\displaystyle \ell ^{p}}$ defined above. For uncountable sets ${\displaystyle I}$ this is a non-separable Banach space which can be seen as the locally convex direct limit of ${\displaystyle \ell ^{p}}$-sequence spaces.^[4]

For ${\displaystyle p=2,}$ the ${\displaystyle \|\,\cdot \,\|_{2}}$-norm is even induced by a canonical inner product ${\displaystyle \langle \,\cdot ,\,\cdot \rangle ,}$ called the Euclidean inner product, which means that ${\displaystyle \|\mathbf {x} \|_{2}={\sqrt {\langle \mathbf {x} ,\mathbf {x} \rangle }}}$ holds for all vectors ${\displaystyle \mathbf {x} .}$ This inner product can expressed in terms of the norm by using the polarization identity. On ${\displaystyle \ell ^{2},}$ it can be defined by $${\displaystyle \langle \left(x_{i}\right)_{i},\left(y_{n}\right)_{i}\rangle _{\ell ^{2}}~=~\sum _{i}x_{i}{\overline {y_{i}}}}$$ while for the space ${\displaystyle L^{2}(X,\mu )}$ associated with a measure space ${\displaystyle (X,\Sigma ,\mu ),}$ which consists of all square-integrable functions, it is $${\displaystyle \langle f,g\rangle _{L^{2}}=\int _{X}f(x){\overline {g(x)}}\,\mathrm {d} x.}$$

Now consider the case ${\displaystyle p=\infty .}$ Define^{[note 1]} $${\displaystyle \ell ^{\infty }(I)=\{x\in \mathbb {K} ^{I}:\sup \operatorname {range} |x|<+\infty \},}$$ where for all ${\displaystyle x}$^{[5][note 2]} $${\displaystyle \|x\|_{\infty }\equiv \inf\{C\in \mathbb {R} _{\geq 0}:|x_{i}|\leq C{\text{ for all }}i\in I\}={\begin{cases}\sup \operatorname {range} |x|&{\text{if }}X\neq \varnothing ,\\0&{\text{if }}X=\varnothing .\end{cases}}}$$

The index set ${\displaystyle I}$ can be turned into a measure space by giving it the discrete σ-algebra and the counting measure. Then the space ${\displaystyle \ell ^{p}(I)}$ is just a special case of the more general ${\displaystyle L^{p}}$-space (defined below).

An ${\displaystyle L^{p}}$ space may be defined as a space of measurable functions for which the ${\displaystyle p}$-th power of the absolute value is Lebesgue integrable, where functions which agree almost everywhere are identified. More generally, let ${\displaystyle (S,\Sigma ,\mu )}$ be a measure space and ${\displaystyle 1\leq p\leq \infty .}$^{[note 3]} When ${\displaystyle p\neq \infty }$, consider the set ${\displaystyle {\mathcal {L}}^{p}(S,\,\mu )}$ of all measurable functions ${\displaystyle f}$ from ${\displaystyle S}$ to ${\displaystyle \mathbb {C} }$ or ${\displaystyle \mathbb {R} }$ whose absolute value raised to the ${\displaystyle p}$-th power has a finite integral, or in symbols: $${\displaystyle \|f\|_{p}~{\stackrel {\scriptscriptstyle {\text{def}}}{=}}~\left(\int _{S}|f|^{p}\;\mathrm {d} \mu \right)^{1/p}<\infty .}$$

To define the set for ${\displaystyle p=\infty ,}$ recall that two functions ${\displaystyle f}$ and ${\displaystyle g}$ defined on ${\displaystyle S}$ are said to be equal almost everywhere, written ${\displaystyle f=g}$ a.e., if the set ${\displaystyle \{s\in S:f(s)\neq g(s)\}}$ is measurable and has measure zero. Similarly, a measurable function ${\displaystyle f}$ (and its absolute value) is bounded (or dominated) almost everywhere by a real number ${\displaystyle C,}$ written ${\displaystyle |f|\leq C}$ a.e., if the (necessarily) measurable set ${\displaystyle \{s\in S:|f(s)|>C\}}$ has measure zero. The space ${\displaystyle {\mathcal {L}}^{\infty }(S,\mu )}$ is the set of all measurable functions ${\displaystyle f}$ that are bounded almost everywhere (by some real ${\displaystyle C}$) and ${\displaystyle \|f\|_{\infty }}$ is defined as the infimum of these bounds: $${\displaystyle \|f\|_{\infty }~{\stackrel {\scriptscriptstyle {\text{def}}}{=}}~\inf\{C\in \mathbb {R} _{\geq 0}:|f(s)|\leq C{\text{ for almost every }}s\}.}$$ When ${\displaystyle \mu (S)\neq 0}$ then this is the same as the essential supremum of the absolute value of ${\displaystyle f}$:^{[note 4]} $${\displaystyle \|f\|_{\infty }~=~{\begin{cases}\operatorname {esssup} |f|&{\text{if }}\mu (S)>0,\\0&{\text{if }}\mu (S)=0.\end{cases}}}$$

For example, if ${\displaystyle f}$ is a measurable function that is equal to ${\displaystyle 0}$ almost everywhere^{[note 5]} then ${\displaystyle \|f\|_{p}=0}$ for every ${\displaystyle p}$ and thus ${\displaystyle f\in {\mathcal {L}}^{p}(S,\,\mu )}$ for all ${\displaystyle p.}$

For every positive ${\displaystyle p,}$ the value under ${\displaystyle \|\,\cdot \,\|_{p}}$ of a measurable function ${\displaystyle f}$ and its absolute value ${\displaystyle |f|:S\to [0,\infty ]}$ are always the same (that is, ${\displaystyle \|f\|_{p}=\||f|\|_{p}}$ for all ${\displaystyle p}$) and so a measurable function belongs to ${\displaystyle {\mathcal {L}}^{p}(S,\,\mu )}$ if and only if its absolute value does. Because of this, many formulas involving ${\displaystyle p}$-norms are stated only for non-negative real-valued functions. Consider for example the identity ${\displaystyle \|f\|_{p}^{r}=\|f^{r}\|_{p/r},}$ which holds whenever ${\displaystyle f\geq 0}$ is measurable, ${\displaystyle r>0}$ is real, and ${\displaystyle 0<p\leq \infty }$ (here ${\displaystyle \infty /r\;{\stackrel {\scriptscriptstyle {\text{def}}}{=}}\;\infty }$ when ${\displaystyle p=\infty }$). The non-negativity requirement ${\displaystyle f\geq 0}$ can be removed by substituting ${\displaystyle |f|}$ in for ${\displaystyle f,}$ which gives ${\displaystyle \|\,|f|\,\|_{p}^{r}=\|\,|f|^{r}\,\|_{p/r}.}$ Note in particular that when ${\displaystyle p=r}$ is finite then the formula ${\displaystyle \|f\|_{p}^{p}=\||f|^{p}\|_{1}}$ relates the ${\displaystyle p}$-norm to the ${\displaystyle 1}$-norm.

Seminormed space of ${\displaystyle p}$-th power integrable functions

Each set of functions ${\displaystyle {\mathcal {L}}^{p}(S,\,\mu )}$ forms a vector space when addition and scalar multiplication are defined pointwise.^{[note 6]} That the sum of two ${\displaystyle p}$-th power integrable functions ${\displaystyle f}$ and ${\displaystyle g}$ is again ${\displaystyle p}$-th power integrable follows from ${\textstyle \|f+g\|_{p}^{p}\leq 2^{p-1}\left(\|f\|_{p}^{p}+\|g\|_{p}^{p}\right),}$^{[proof 1]} although it is also a consequence of Minkowski's inequality $${\displaystyle \|f+g\|_{p}\leq \|f\|_{p}+\|g\|_{p}}$$ which establishes that ${\displaystyle \|\cdot \|_{p}}$ satisfies the triangle inequality for ${\displaystyle 1\leq p\leq \infty }$ (the triangle inequality does not hold for ${\displaystyle 0<p<1}$). That ${\displaystyle {\mathcal {L}}^{p}(S,\,\mu )}$ is closed under scalar multiplication is due to ${\displaystyle \|\cdot \|_{p}}$ being absolutely homogeneous, which means that ${\displaystyle \|sf\|_{p}=|s|\|f\|_{p}}$ for every scalar ${\displaystyle s}$ and every function ${\displaystyle f.}$

Absolute homogeneity, the triangle inequality, and non-negativity are the defining properties of a seminorm. Thus ${\displaystyle \|\cdot \|_{p}}$ is a seminorm and the set ${\displaystyle {\mathcal {L}}^{p}(S,\,\mu )}$ of ${\displaystyle p}$-th power integrable functions together with the function ${\displaystyle \|\cdot \|_{p}}$ defines a seminormed vector space. In general, the seminorm ${\displaystyle \|\cdot \|_{p}}$ is not a norm because there might exist measurable functions ${\displaystyle f}$ that satisfy ${\displaystyle \|f\|_{p}=0}$ but are not identically equal to ${\displaystyle 0}$^{[note 5]} (${\displaystyle \|\cdot \|_{p}}$ is a norm if and only if no such ${\displaystyle f}$ exists).

Zero sets of ${\displaystyle p}$-seminorms

If ${\displaystyle f}$ is measurable and equals ${\displaystyle 0}$ a.e. then ${\displaystyle \|f\|_{p}=0}$ for all positive ${\displaystyle p\leq \infty .}$ On the other hand, if ${\displaystyle f}$ is a measurable function for which there exists some ${\displaystyle 0<p\leq \infty }$ such that ${\displaystyle \|f\|_{p}=0}$ then ${\displaystyle f=0}$ almost everywhere. When ${\displaystyle p}$ is finite then this follows from the ${\displaystyle p=1}$ case and the formula ${\displaystyle \|f\|_{p}^{p}=\||f|^{p}\|_{1}}$ mentioned above.

Thus if ${\displaystyle p\leq \infty }$ is positive and ${\displaystyle f}$ is any measurable function, then ${\displaystyle \|f\|_{p}=0}$ if and only if ${\displaystyle f=0}$ almost everywhere. Since the right hand side (${\displaystyle f=0}$ a.e.) does not mention ${\displaystyle p,}$ it follows that all ${\displaystyle \|\cdot \|_{p}}$ have the same zero set (it does not depend on ${\displaystyle p}$). So denote this common set by $${\displaystyle {\mathcal {N}}\;{\stackrel {\scriptscriptstyle {\text{def}}}{=}}\;\{f:f=0\ \mu {\text{-almost everywhere}}\}=\{f\in {\mathcal {L}}^{p}(S,\,\mu ):\|f\|_{p}=0\}\qquad \forall \ p.}$$ This set is a vector subspace of ${\displaystyle {\mathcal {L}}^{p}(S,\,\mu )}$ for every positive ${\displaystyle p\leq \infty .}$

Quotient vector space

Like every seminorm, the seminorm ${\displaystyle \|\cdot \|_{p}}$ induces a norm (defined shortly) on the canonical quotient vector space of ${\displaystyle {\mathcal {L}}^{p}(S,\,\mu )}$ by its vector subspace ${\textstyle {\mathcal {N}}=\{f\in {\mathcal {L}}^{p}(S,\,\mu ):\|f\|_{p}=0\}.}$ This normed quotient space is called Lebesgue space and it is the subject of this article. We begin by defining the quotient vector space.

Given any ${\displaystyle f\in {\mathcal {L}}^{p}(S,\,\mu ),}$ the coset ${\displaystyle f+{\mathcal {N}}\;{\stackrel {\scriptscriptstyle {\text{def}}}{=}}\;\{f+h:h\in {\mathcal {N}}\}}$ consists of all measurable functions ${\displaystyle g}$ that are equal to ${\displaystyle f}$ almost everywhere. The set of all cosets, typically denoted by $${\displaystyle {\mathcal {L}}^{p}(S,\mu )/{\mathcal {N}}~~{\stackrel {\scriptscriptstyle {\text{def}}}{=}}~~\{f+{\mathcal {N}}:f\in {\mathcal {L}}^{p}(S,\mu )\},}$$ forms a vector space with origin ${\displaystyle 0+{\mathcal {N}}={\mathcal {N}}}$ when vector addition and scalar multiplication are defined by ${\displaystyle (f+{\mathcal {N}})+(g+{\mathcal {N}})\;{\stackrel {\scriptscriptstyle {\text{def}}}{=}}\;(f+g)+{\mathcal {N}}}$ and ${\displaystyle s(f+{\mathcal {N}})\;{\stackrel {\scriptscriptstyle {\text{def}}}{=}}\;(sf)+{\mathcal {N}}.}$ This particular quotient vector space will be denoted by ${\displaystyle L^{p}(S,\,\mu )~{\stackrel {\scriptscriptstyle {\text{def}}}{=}}~{\mathcal {L}}^{p}(S,\mu )/{\mathcal {N}}.}$

Two cosets are equal ${\displaystyle f+{\mathcal {N}}=g+{\mathcal {N}}}$ if and only if ${\displaystyle g\in f+{\mathcal {N}}}$ (or equivalently, ${\displaystyle f-g\in {\mathcal {N}}}$), which happens if and only if ${\displaystyle f=g}$ almost everywhere; if this is the case then ${\displaystyle f}$ and ${\displaystyle g}$ are identified in the quotient space.

The ${\displaystyle p}$-norm on the quotient vector space

Given any ${\displaystyle f\in {\mathcal {L}}^{p}(S,\,\mu ),}$ the value of the seminorm ${\displaystyle \|\cdot \|_{p}}$ on the coset ${\displaystyle f+{\mathcal {N}}=\{f+h:h\in {\mathcal {N}}\}}$ is constant and equal to ${\displaystyle \|f\|_{p};}$ denote this unique value by ${\displaystyle \|f+{\mathcal {N}}\|_{p},}$ so that: $${\displaystyle \|f+{\mathcal {N}}\|_{p}\;{\stackrel {\scriptscriptstyle {\text{def}}}{=}}\;\|f\|_{p}.}$$ This assignment ${\displaystyle f+{\mathcal {N}}\mapsto \|f+{\mathcal {N}}\|_{p}}$ defines a map, which will also be denoted by ${\displaystyle \|\cdot \|_{p},}$ on the quotient vector space $${\displaystyle L^{p}(S,\mu )~~{\stackrel {\scriptscriptstyle {\text{def}}}{=}}~~{\mathcal {L}}^{p}(S,\mu )/{\mathcal {N}}~=~\{f+{\mathcal {N}}:f\in {\mathcal {L}}^{p}(S,\mu )\}.}$$ This map is a norm on ${\displaystyle L^{p}(S,\mu )}$ called the ${\displaystyle p}$-norm. The value ${\displaystyle \|f+{\mathcal {N}}\|_{p}}$ of a coset ${\displaystyle f+{\mathcal {N}}}$ is independent of the particular function ${\displaystyle f}$ that was chosen to represent the coset, meaning that if ${\displaystyle {\mathcal {C}}\in L^{p}(S,\mu )}$ is any coset then ${\displaystyle \|{\mathcal {C}}\|_{p}=\|f\|_{p}}$ for every ${\displaystyle f\in {\mathcal {C}}}$ (since ${\displaystyle {\mathcal {C}}=f+{\mathcal {N}}}$ for every ${\displaystyle f\in {\mathcal {C}}}$).

The Lebesgue ${\displaystyle L^{p}}$ space

The normed vector space ${\displaystyle \left(L^{p}(S,\mu ),\|\cdot \|_{p}\right)}$ is called ${\displaystyle L^{p}}$ space or the Lebesgue space of ${\displaystyle p}$-th power integrable functions and it is a Banach space for every ${\displaystyle 1\leq p\leq \infty }$ (meaning that it is a complete metric space, a result that is sometimes called the Riesz–Fischer theorem). When the underlying measure space ${\displaystyle S}$ is understood then ${\displaystyle L^{p}(S,\mu )}$ is often abbreviated ${\displaystyle L^{p}(\mu ),}$ or even just ${\displaystyle L^{p}.}$ Depending on the author, the subscript notation ${\displaystyle L_{p}}$ might denote either ${\displaystyle L^{p}(S,\mu )}$ or ${\displaystyle L^{1/p}(S,\mu ).}$

If the seminorm ${\displaystyle \|\cdot \|_{p}}$ on ${\displaystyle {\mathcal {L}}^{p}(S,\,\mu )}$ happens to be a norm (which happens if and only if ${\displaystyle {\mathcal {N}}=\{0\}}$) then the normed space ${\displaystyle \left({\mathcal {L}}^{p}(S,\,\mu ),\|\cdot \|_{p}\right)}$ will be linearly isometrically isomorphic to the normed quotient space ${\displaystyle \left(L^{p}(S,\mu ),\|\cdot \|_{p}\right)}$ via the canonical map ${\displaystyle g\in {\mathcal {L}}^{p}(S,\,\mu )\mapsto \{g\}}$ (since ${\displaystyle g+{\mathcal {N}}=\{g\}}$); in other words, they will be, up to a linear isometry, the same normed space and so they may both be called "${\displaystyle L^{p}}$ space".

The above definitions generalize to Bochner spaces.

In general, this process cannot be reversed: there is no consistent way to define a "canonical" representative of each coset of ${\displaystyle {\mathcal {N}}}$ in ${\displaystyle L^{p}.}$ For ${\displaystyle L^{\infty },}$ however, there is a theory of lifts enabling such recovery.

Similar to the ${\displaystyle \ell ^{p}}$ spaces, ${\displaystyle L^{2}}$ is the only Hilbert space among ${\displaystyle L^{p}}$ spaces. In the complex case, the inner product on ${\displaystyle L^{2}}$ is defined by $${\displaystyle \langle f,g\rangle =\int _{S}f(x){\overline {g(x)}}\,\mathrm {d} \mu (x)}$$

The additional inner product structure allows for a richer theory, with applications to, for instance, Fourier series and quantum mechanics. Functions in ${\displaystyle L^{2}}$ are sometimes called square-integrable functions, quadratically integrable functions or square-summable functions, but sometimes these terms are reserved for functions that are square-integrable in some other sense, such as in the sense of a Riemann integral (Titchmarsh 1976).

If we use complex-valued functions, the space ${\displaystyle L^{\infty }}$ is a commutative C*-algebra with pointwise multiplication and conjugation. For many measure spaces, including all sigma-finite ones, it is in fact a commutative von Neumann algebra. An element of ${\displaystyle L^{\infty }}$ defines a bounded operator on any ${\displaystyle L^{p}}$ space by multiplication.

For ${\displaystyle 1\leq p\leq \infty }$ the ${\displaystyle \ell ^{p}}$ spaces are a special case of ${\displaystyle L^{p}}$ spaces, when ${\displaystyle S=\mathbb {N} )}$ consists of the natural numbers and ${\displaystyle \mu }$ is the counting measure on ${\displaystyle \mathbb {N} .}$ More generally, if one considers any set ${\displaystyle S}$ with the counting measure, the resulting ${\displaystyle L^{p}}$ space is denoted ${\displaystyle \ell ^{p}(S).}$ For example, the space ${\displaystyle \ell ^{p}(\mathbb {Z} )}$ is the space of all sequences indexed by the integers, and when defining the ${\displaystyle p}$-norm on such a space, one sums over all the integers. The space ${\displaystyle \ell ^{p}(n),}$ where ${\displaystyle n}$ is the set with ${\displaystyle n}$ elements, is ${\displaystyle \mathbb {R} ^{n}}$ with its ${\displaystyle p}$-norm as defined above. As any Hilbert space, every space ${\displaystyle L^{2}}$ is linearly isometric to a suitable ${\displaystyle \ell ^{2}(I),}$ where the cardinality of the set ${\displaystyle I}$ is the cardinality of an arbitrary Hilbertian basis for this particular ${\displaystyle L^{2}.}$

As in the discrete case, if there exists ${\displaystyle q<\infty }$ such that ${\displaystyle f\in L^{\infty }(S,\mu )\cap L^{q}(S,\mu ),}$ then^{[citation needed]} $${\displaystyle \|f\|_{\infty }=\lim _{p\to \infty }\|f\|_{p}.}$$

Hölder's inequality

Suppose ${\displaystyle p,q,r\in [1,\infty ]}$ satisfy ${\displaystyle {\tfrac {1}{p}}+{\tfrac {1}{q}}={\tfrac {1}{r}}}$ (where ${\displaystyle {\tfrac {1}{\infty }}~{\stackrel {\scriptscriptstyle {\text{def}}}{=}}~0}$). If ${\displaystyle f\in L^{p}(S,\mu )}$ and ${\displaystyle g\in L^{q}(S,\mu )}$ then ${\displaystyle fg\in L^{r}(S,\mu )}$ and^[6] $${\displaystyle \|fg\|_{r}~\leq ~\|f\|_{p}\,\|g\|_{q}.}$$

This inequality, called Hölder's inequality, is in some sense optimal^[6] since if ${\displaystyle r=1}$ (so ${\displaystyle {\tfrac {1}{p}}+{\tfrac {1}{q}}=1}$) and ${\displaystyle f}$ is a measurable function such that $${\displaystyle \sup _{\|g\|_{q}\leq 1}\,\int _{S}|fg|\,\mathrm {d} \mu ~<~\infty }$$ where the supremum is taken over the closed unit ball of ${\displaystyle L^{q}(S,\mu ),}$ then ${\displaystyle f\in L^{p}(S,\mu )}$ and $${\displaystyle \|f\|_{p}~=~\sup _{\|g\|_{q}\leq 1}\,\int _{S}fg\,\mathrm {d} \mu .}$$

Minkowski inequality

Minkowski inequality, which states that ${\displaystyle \|\cdot \|_{p}}$ satisfies the triangle inequality, can be generalized: If the measurable function ${\displaystyle F:M\times N\to \mathbb {R} }$ is non-negative (where ${\displaystyle (M,\mu )}$ and ${\displaystyle (N,\nu )}$ are measure spaces) then for all ${\displaystyle 1\leq p\leq q\leq \infty ,}$^[7] $${\displaystyle \left\|\left\|F(\,\cdot ,n)\right\|_{L^{p}(M,\mu )}\right\|_{L^{q}(N,\nu )}~\leq ~\left\|\left\|F(m,\cdot )\right\|_{L^{q}(N,\nu )}\right\|_{L^{p}(M,\mu )}\ .}$$

If ${\displaystyle 1\leq p<\infty }$ then every non-negative ${\displaystyle f\in L^{p}(\mu )}$ has an atomic decomposition,^[8] meaning that there exist a sequence ${\displaystyle (r_{n})_{n\in \mathbb {Z} }}$ of non-negative real numbers and a sequence of non-negative functions ${\displaystyle (f_{n})_{n\in \mathbb {Z} },}$ called the atoms, whose supports ${\displaystyle \left(\operatorname {supp} f_{n}\right)_{n\in \mathbb {Z} }}$ are pairwise disjoint sets of measure ${\displaystyle \mu \left(\operatorname {supp} f_{n}\right)\leq 2^{n+1},}$ such that $${\displaystyle f~=~\sum _{n\in \mathbb {Z} }r_{n}\,f_{n}\,,}$$ and for every integer ${\displaystyle n\in \mathbb {Z} ,}$ $${\displaystyle \|f_{n}\|_{\infty }~\leq ~2^{-{\tfrac {n}{p}}}\,,}$$ and $${\displaystyle {\tfrac {1}{2}}\|f\|_{p}^{p}~\leq ~\sum _{n\in \mathbb {Z} }r_{n}^{p}~\leq ~2\|f\|_{p}^{p}\,,}$$ and where moreover, the sequence of functions ${\displaystyle (r_{n}f_{n})_{n\in \mathbb {Z} }}$ depends only on ${\displaystyle f}$ (it is independent of ${\displaystyle p}$).^[8] These inequalities guarantee that ${\displaystyle \|f_{n}\|_{p}^{p}\leq 2}$ for all integers ${\displaystyle n}$ while the supports of ${\displaystyle (f_{n})_{n\in \mathbb {Z} }}$ being pairwise disjoint implies^[8] $${\displaystyle \|f\|_{p}^{p}~=~\sum _{n\in \mathbb {Z} }r_{n}^{p}\,\|f_{n}\|_{p}^{p}\,.}$$

An atomic decomposition can be explicitly given by first defining for every integer ${\displaystyle n\in \mathbb {Z} ,}$^[8] $${\displaystyle t_{n}=\inf\{t\in \mathbb {R} :\mu (f>t)<2^{n}\}}$$ (this infimum is attained by ${\displaystyle t_{n};}$ that is, ${\displaystyle \mu (f>t_{n})<2^{n}}$ holds) and then letting $${\displaystyle r_{n}~=~2^{n/p}\,t_{n}~{\text{ and }}\quad f_{n}~=~{\frac {f}{r_{n}}}\,\mathbf {1} _{(t_{n+1}<f\leq t_{n})}}$$ where ${\displaystyle \mu (f>t)=\mu (\{s:f(s)>t\})}$ denotes the measure of the set ${\displaystyle (f>t):=\{s\in S:f(s)>t\}}$ and ${\displaystyle \mathbf {1} _{(t_{n+1}<f\leq t_{n})}}$ denotes the indicator function of the set ${\displaystyle (t_{n+1}<f\leq t_{n}):=\{s\in S:t_{n+1}<f(s)\leq t_{n}\}.}$ The sequence ${\displaystyle (t_{n})_{n\in \mathbb {Z} }}$ is decreasing and converges to ${\displaystyle 0}$ as ${\displaystyle n\to \infty .}$^[8] Consequently, if ${\displaystyle t_{n}=0}$ then ${\displaystyle t_{n+1}=0}$ and ${\displaystyle (t_{n+1}<f\leq t_{n})=\varnothing }$ so that ${\displaystyle f_{n}={\frac {1}{r_{n}}}\,f\,\mathbf {1} _{(t_{n+1}<f\leq t_{n})}}$ is identically equal to ${\displaystyle 0}$ (in particular, the division ${\displaystyle {\tfrac {1}{r_{n}}}}$ by ${\displaystyle r_{n}=0}$ causes no issues).