Content deleted Content added
Fishsicles (talk | contribs) m →Formulation: Double-checked the hyperfunction formulations against each other - they're not equivalent at H(0) |
mNo edit summary |
||
| (9 intermediate revisions by 7 users not shown) | |||
Line 16:
==Formulation==
Taking the convention that {{math|''H''(0) {{=}} 1}}, the Heaviside function may be defined as:
*
*
*
* a [[hyperfunction]] <math display="block">H(x) =: \left(1-\frac{1}{2\pi i}\log z,\ -\frac{1}{2\pi i}\log z\right)</math> where {{math|log ''z''}} is the [[Complex logarithm#Principal value|principal value of the complex logarithm]] of {{mvar|z}}▼
For the alternative convention that {{math|''H''(0) {{=}} {{sfrac|1|2}}}}, it may be expressed as:
*
*
*
*
▲*
Other definitions which are undefined at {{math|''H''(0)}} include:
*
* The derivative of the [[ramp function]]: <math display="block">H(x) := \frac{d}{dx} \max \{ x, 0 \}\quad \mbox{for } x \ne 0</math>
* Expressed in terms of the [[absolute value]] function, such as:<math display="block"> H(x) = \frac{x + |x|}{2x}</math>
==Relationship with Dirac delta==
The [[Dirac delta function]] is the [[weak derivative]] of the Heaviside function:<math display="block">\delta(x)= \frac{d}{dx} \ H(x),</math>Hence the Heaviside function can be considered to be the [[integral]] of the Dirac delta function. This is sometimes written as:<math display="block">H(x) := \int_{-\infty}^x \delta(s)\,ds,</math>although this expansion may not hold (or even make sense) for {{math|''x'' {{=}} 0}}, depending on which formalism one uses to give meaning to integrals involving {{mvar|δ}}. In this context, the Heaviside function is the [[cumulative distribution function]] of a [[random variable]] which is [[almost surely]] 0. (See [[Constant random variable]].)▼
▲although this expansion may not hold (or even make sense) for {{math|''x'' {{=}} 0}}, depending on which formalism one uses to give meaning to integrals involving {{mvar|δ}}. In this context, the Heaviside function is the [[cumulative distribution function]] of a [[random variable]] which is [[almost surely]] 0. (See [[Constant random variable]].)
== Analytic approximations ==
Approximations to the Heaviside step function are of use in [[biochemistry]] and [[neuroscience]], where [[logistic function|logistic]] approximations of step functions (such as the [[Hill equation (biochemistry)|Hill]] and the [[Michaelis–Menten kinetics|Michaelis–Menten equations]]) may be used to approximate binary cellular switches in response to chemical signals.
For a [[
[[File:Step function approximation.png|alt=A set of functions that successively approach the step function|thumb|500x500px|<math>\tfrac{1}{2} + \tfrac{1}{2} \tanh(kx) = \frac{1}{1+e^{-2kx}}</math><br>approaches the step function as {{math|''k'' → ∞}}.|none]]There are [[Sigmoid function#Examples|many other smooth, analytic approximations]] to the step function.<ref>{{MathWorld | urlname=HeavisideStepFunction | title=Heaviside Step Function}}</ref> Among the possibilities are:<math display="block">\begin{align}
H(x) &= \lim_{k \to \infty} \left(\tfrac{1}{2} + \tfrac{1}{\pi}\arctan kx\right)\\
H(x) &= \lim_{k \to \infty}\left(\tfrac{1}{2} + \tfrac12\operatorname{erf} kx\right)
\end{align}</math>These limits hold [[pointwise]] and in the sense of [[distribution (mathematics)|distributions]]. In general, however, [[pointwise convergence]] need not imply distributional convergence, and vice versa distributional convergence need not imply pointwise convergence. (However, if all members of a pointwise convergent sequence of functions are uniformly bounded by some "nice" function, then [[Lebesgue dominated convergence theorem|convergence holds in the sense of distributions too]].)▼
▲These limits hold [[pointwise]] and in the sense of [[distribution (mathematics)|distributions]]. In general, however, pointwise convergence need not imply distributional convergence, and vice versa distributional convergence need not imply pointwise convergence. (However, if all members of a pointwise convergent sequence of functions are uniformly bounded by some "nice" function, then [[Lebesgue dominated convergence theorem|convergence holds in the sense of distributions too]].)
In general, any [[cumulative distribution function]] of a [[continuous distribution|continuous]] [[probability distribution]] that is peaked around zero and has a parameter that controls for [[variance]] can serve as an approximation, in the limit as the variance approaches zero. For example, all three of the above approximations are [[cumulative distribution function|cumulative distribution functions]] of common probability distributions: the [[logistic distribution|logistic]], [[Cauchy distribution|Cauchy]] and [[normal distribution|normal]] distributions, respectively.
== Non-Analytic approximations ==
Approximations to the Heaviside step function could be made through [[Non-analytic_smooth_function#Smooth_transition_functions|Smooth transition function]] like <math> 1 \leq m \to \infty </math>:<math display="block">\begin{align}f(x) &= \begin{cases}
{\displaystyle
\frac{1}{2}\left(1+\tanh\left(m\frac{2x}{1-x^2}\right)\right)}, & |x| < 1 \\
Line 70 ⟶ 58:
==Integral representations==
Often an [[integration (mathematics)|integral]] representation of the Heaviside step function is useful:<math display="block">\begin{align}
H(x)&=\lim_{ \varepsilon \to 0^+} -\frac{1}{2\pi i}\int_{-\infty}^\infty \frac{1}{\tau+i\varepsilon} e^{-i x \tau} d\tau \\
&=\lim_{ \varepsilon \to 0^+} \ \frac{1}{2\pi i}\int_{-\infty}^\infty \frac{1}{\tau-i\varepsilon} e^{i x \tau} d\tau
\end{align}</math>where the second representation is easy to deduce from the first, given that the step function is real and thus is its own [[complex conjugate]].▼
▲where the second representation is easy to deduce from the first, given that the step function is real and thus is its own complex conjugate.
== Zero argument ==
Since {{mvar|H}} is usually used in [[Integral|integration]], and the value of a function at a single point does not affect its integral, it rarely matters what particular value is chosen of {{math|''H''(0)}}. Indeed when {{mvar|H}} is considered as a [[distribution (mathematics)|distribution]] or an element of {{math|''L''{{isup|∞}}}} (see [[Lp space|{{math|''L{{isup|p}}''}} space]]) it does not even make sense to talk of a value at zero, since such objects are only defined [[almost everywhere]]. If using some analytic approximation (as in the [[#Analytic approximations|examples above]]) then often whatever happens to be the relevant limit at zero is used.
There exist various reasons for choosing a particular value.
* {{math|''H''(0) {{=}} {{sfrac|1|2}}}} is often used since the [[graph of a function|graph]] then has [[rotational symmetry]]; put another way, {{math|''H'' − {{sfrac|1|2}}}} is then an [[odd function]]. In this case the following relation with the [[sign function]] holds for all {{mvar|x}}: <math display="block">
* {{math|''H''(0) {{=}} 1}} is used when {{mvar|H}} needs to be [[right-continuous]]. For instance [[cumulative distribution function]]s are usually taken to be right continuous, as are functions integrated against in [[Lebesgue–Stieltjes integration]]. In this case {{mvar|H}} is the [[indicator function]] of a [[closed set|closed]] semi-infinite interval: <math display="block"> H(x) = \mathbf{1}_{[0,\infty)}(x).</math> The corresponding probability distribution is the [[degenerate distribution]].
* {{math|''H''(0) {{=}} 0}} is used when {{mvar|H}} needs to be [[left-continuous]]. In this case {{mvar|H}} is an indicator function of an [[open set|open]] semi-infinite interval: <math display="block"> H(x) = \mathbf{1}_{(0,\infty)}(x).</math>
* In functional-analysis contexts from [[optimization]] and [[game theory]], it is often useful to define the Heaviside function as a [[Multivalued function|set-valued function]] to preserve the continuity of the limiting functions and ensure the existence of certain solutions. In these cases, the Heaviside function returns a whole interval of possible solutions, {{math|''H''(0) {{=}} [0,1]}}.
==Discrete form==
An alternative form of the unit step, defined instead as a function <math>H : \mathbb{Z} \rarr \mathbb{R}</math> (that is, taking in a discrete variable {{mvar|n}}), is:<math display="block">H[n]=\begin{cases} 0, & n < 0, \\ 1, & n \ge 0, \end{cases} </math>Or using the half-maximum convention:<ref>{{cite book |last=Bracewell |first=Ronald Newbold |date=2000 |title=The Fourier transform and its applications |language=en |___location=New York |publisher=McGraw-Hill |isbn=0-07-303938-1 |page=61 |edition=3rd}}</ref><math display="block">H[n]=\begin{cases} 0, & n < 0, \\ \tfrac12, & n = 0,\\ 1, & n > 0, \end{cases} </math>where {{mvar|n}} is an [[integer]]. If {{mvar|n}} is an integer, then {{math|''n'' < 0}} must imply that {{math|''n'' ≤ −1}}, while {{math|''n'' > 0}} must imply that the function attains unity at {{math|1=''n'' = 1}}. Therefore the "step function" exhibits ramp-like behavior over the ___domain of {{closed-closed|−1, 1}}, and cannot authentically be a step function, using the half-maximum convention.
Unlike the continuous case, the definition of {{math|''H''[0]}} is significant.
The discrete-time unit impulse is the first difference of the discrete-time step:<math display="block"> \delta[n] = H[n] - H[n-1].</math>This function is the cumulative summation of the [[Kronecker delta]]:<math display="block"> H[n] = \sum_{k=-\infty}^{n} \delta[k], </math>where <math display="inline"> \delta[k] = \delta_{k,0} </math> is the [[degenerate distribution|discrete unit impulse function]].
== Antiderivative and derivative==
The [[ramp function]] is an [[antiderivative]] of the Heaviside step function:<math display="block">\int_{-\infty}^{x} H(\xi)\,d\xi = x H(x) = \max\{0,x\} \,.</math>The [[distributional derivative]] of the Heaviside step function is the [[Dirac delta function]]:<math display="block"> \frac{d H(x)}{dx} = \delta(x) \,.</math>
== Fourier transform ==
The [[Fourier transform]] of the Heaviside step function is a distribution. Using one choice of constants for the definition of the Fourier transform we have
<math display="block">\hat{H}(s) = \lim_{N\to\infty}\int^N_{-N} e^{-2\pi i x s} H(x)\,dx = \frac{1}{2} \left( \delta(s) - \frac{i}{\pi} \operatorname{p.v.}\frac{1}{s} \right).</math>Here {{math|p.v.{{sfrac|1|''s''}}}} is the [[distribution (mathematics)|distribution]] that takes a test function {{mvar|φ}} to the [[Cauchy principal value]] of <math>\textstyle\int_{-\infty}^\infty \frac{\varphi(s)}{s} \, ds</math>. The limit appearing in the integral is also taken in the sense of (tempered) distributions.
== Unilateral Laplace transform ==
The [[Laplace transform]] of the Heaviside step function is a [[meromorphic function]]. Using the unilateral Laplace transform we have:<math display="block">\begin{align}
\hat{H}(s) &= \lim_{N\to\infty}\int^N_{0} e^{-sx} H(x)\,dx\\
&= \lim_{N\to\infty}\int^N_{0} e^{-sx} \,dx\\
&= \frac{1}{s} \end{align}</math>When the [[Laplace transform#Bilateral Laplace transform|bilateral transform]] is used, the integral can be split in two parts and the result will be the same.▼
▲When the bilateral transform is used, the integral can be split in two parts and the result will be the same.
==See also==
{{Div col|colwidth=25em}}
* [[Gamma function]]
* [[Dirac delta function]]
* [[Indicator function]]
| |||