From 62759ea3f910fae2617d033bf8f878d7574f4edd Mon Sep 17 00:00:00 2001 From: Prefetch Date: Sun, 7 Nov 2021 19:34:18 +0100 Subject: Expand knowledge base, reorganize measure theory, update gitignore --- .../concept/gronwall-bellman-inequality/index.pdc | 210 +++++++++++++++++++++ content/know/concept/ito-calculus/index.pdc | 180 ++++++++++++++++-- content/know/concept/ito-integral/index.pdc | 7 +- content/know/concept/martingale/index.pdc | 20 +- content/know/concept/random-variable/index.pdc | 40 +++- content/know/concept/sigma-algebra/index.pdc | 61 ------ content/know/concept/stochastic-process/index.pdc | 62 ++++++ content/know/concept/wiener-process/index.pdc | 32 +--- .../know/concept/young-dupre-relation/index.pdc | 2 +- 9 files changed, 492 insertions(+), 122 deletions(-) create mode 100644 content/know/concept/gronwall-bellman-inequality/index.pdc create mode 100644 content/know/concept/stochastic-process/index.pdc (limited to 'content/know') diff --git a/content/know/concept/gronwall-bellman-inequality/index.pdc b/content/know/concept/gronwall-bellman-inequality/index.pdc new file mode 100644 index 0000000..1f093ae --- /dev/null +++ b/content/know/concept/gronwall-bellman-inequality/index.pdc @@ -0,0 +1,210 @@ +--- +title: "Grönwall-Bellman inequality" +firstLetter: "G" +publishDate: 2021-11-07 +categories: +- Mathematics + +date: 2021-11-07T09:51:57+01:00 +draft: false +markup: pandoc +--- + +# Grönwall-Bellman inequality + +Suppose we have a first-order ordinary differential equation +for some function $u(t)$, and that it can be shown from this equation +that the derivative $u'(t)$ is bounded as follows: + +$$\begin{aligned} + u'(t) + \le \beta(t) \: u(t) +\end{aligned}$$ + +Where $\beta(t)$ is known. +Then **Grönwall's inequality** states that the solution $u(t)$ is bounded: + +$$\begin{aligned} + \boxed{ + u(t) + \le u(0) \exp\!\bigg( \int_0^t \beta(s) \dd{s} \bigg) + } +\end{aligned}$$ + +
+ + + +
+ +Grönwall's inequality can be generalized to non-differentiable functions. +Suppose we know: + +$$\begin{aligned} + u(t) + \le \alpha(t) + \int_0^t \beta(s) \: u(s) \dd{s} +\end{aligned}$$ + +Where $\alpha(t)$ and $\beta(t)$ are known. +Then the **Grönwall-Bellman inequality** states that: + +$$\begin{aligned} + \boxed{ + u(t) + \le \alpha(t) + \int_0^t \alpha(s) \: \beta(s) \exp\!\bigg( \int_s^t \beta(r) \dd{r} \bigg) \dd{s} + } +\end{aligned}$$ + +
+ + + +
+ +In the special case where $\alpha(t)$ is non-decreasing with $t$, +the inequality reduces to: + +$$\begin{aligned} + \boxed{ + u(t) + \le \alpha(t) \exp\!\bigg( \int_0^t \beta(s) \dd{s} \bigg) + } +\end{aligned}$$ + +
+ + + +
+ + + +## References +1. U.H. Thygesen, + *Lecture notes on diffusions and stochastic differential equations*, + 2021, Polyteknisk Kompendie. diff --git a/content/know/concept/ito-calculus/index.pdc b/content/know/concept/ito-calculus/index.pdc index 576e09a..3527b1d 100644 --- a/content/know/concept/ito-calculus/index.pdc +++ b/content/know/concept/ito-calculus/index.pdc @@ -12,10 +12,10 @@ markup: pandoc # Itō calculus -Given two time-indexed [random variables](/know/concept/random-variable/) -(i.e. stochastic processes) $F_t$ and $G_t$, -then consider the following random variable $X_t$, -where $B_t$ is the [Wiener process](/know/concept/wiener-process/): +Given two [stochastic processes](/know/concept/stochastic-process/) +$F_t$ and $G_t$, consider the following random variable $X_t$, +where $B_t$ is the [Wiener process](/know/concept/wiener-process/), +i.e. Brownian motion: $$\begin{aligned} X_t @@ -27,7 +27,7 @@ assuming $G_t$ is Itō-integrable. We call $X_t$ an **Itō process** if $F_t$ is locally integrable, and the initial condition $X_0$ is known, i.e. $X_0$ is $\mathcal{F}_0$-measurable, -where $\mathcal{F}_t$ is the [filtration](/know/concept/sigma-algebra/) +where $\mathcal{F}_t$ is the filtration to which $F_t$, $G_t$ and $B_t$ are adapted. The above definition of $X_t$ is often abbreviated as follows, where $X_0$ is implicit: @@ -39,8 +39,18 @@ $$\begin{aligned} Typically, $F_t$ is referred to as the **drift** of $X_t$, and $G_t$ as its **intensity**. +Because the Itō integral of $G_t$ is a +[martingale](/know/concept/martingale/), +it does not contribute to the mean of $X_t$: + +$$\begin{aligned} + \mathbf{E}[X_t] + = \int_0^t \mathbf{E}[F_s] \dd{s} +\end{aligned}$$ + Now, consider the following **Itō stochastic differential equation** (SDE), -where $\xi_t = \dv*{B_t}{t}$ is white noise: +where $\xi_t = \dv*{B_t}{t}$ is white noise, +informally treated as the $t$-derivative of $B_t$: $$\begin{aligned} \dv{X_t}{t} @@ -51,15 +61,6 @@ An Itō process $X_t$ is said to satisfy this equation if $f(X_t, t) = F_t$ and $g(X_t, t) = G_t$, in which case $X_t$ is also called an **Itō diffusion**. -Because the Itō integral of $G_t$ is a -[martingale](/know/concept/martingale/), -it does not contribute to the mean of $X_t$: - -$$\begin{aligned} - \mathbf{E}[X_t] - = \int_0^t \mathbf{E}[F_s] \dd{s} -\end{aligned}$$ - ## Itō's lemma @@ -204,9 +205,156 @@ $$\begin{aligned} 0 &= f(x) \: h'(x) + \frac{1}{2} g^2(x) \: h''(x) \\ - &= \Big( f(x) - \frac{1}{2} g^2(x) \frac{2 f(x)}{g(x)} \Big) \exp\!\bigg( \!-\!\! \int_{x_1}^x \frac{2 f(y)}{g^2(y)} \dd{y} \bigg) + &= \Big( f(x) - \frac{1}{2} g^2(x) \frac{2 f(x)}{g^2(x)} \Big) \exp\!\bigg( \!-\!\! \int_{x_1}^x \frac{2 f(y)}{g^2(y)} \dd{y} \bigg) +\end{aligned}$$ + + +## Existence and uniqueness + +It is worth knowing under what condition a solution to a given SDE exists, +in the sense that it is finite on the entire time axis. +Suppose the drift $f$ and intensity $g$ satisfy these inequalities, +for some known constant $K$ and for all $x$: + +$$\begin{aligned} + x f(x) \le K (1 + x^2) + \qquad \quad + g^2(x) \le K (1 + x^2) +\end{aligned}$$ + +When this is satisfied, we can find the following upper bound +on an Itō process $X_t$, +which clearly implies that $X_t$ is finite for all $t$: + +$$\begin{aligned} + \boxed{ + \mathbf{E}[X_t^2] + \le \big(X_0^2 + 3 K t\big) \exp\!\big(3 K t\big) + } +\end{aligned}$$ + +
+ + + +
+ +If a solution exists, it is also worth knowing whether it is unique. +Suppose that $f$ and $g$ satisfy the following inequalities, +for some constant $K$ and for all $x$ and $y$: + +$$\begin{aligned} + \big| f(x) - f(y) \big| \le K \big| x - y \big| + \qquad \quad + \big| g(x) - g(y) \big| \le K \big| x - y \big| +\end{aligned}$$ + +Let $X_t$ and $Y_t$ both be solutions to a given SDE, +but the initial conditions need not be the same, +such that the difference is initially $X_0 \!-\! Y_0$. +Then the difference $X_t \!-\! Y_t$ is bounded by: + +$$\begin{aligned} + \boxed{ + \mathbf{E}\big[ (X_t - Y_t)^2 \big] + \le (X_0 - Y_0)^2 \exp\!\Big( \big(2 K \!+\! K^2 \big) t \Big) + } +\end{aligned}$$ + +
+ + + +
+ +Using these properties, it can then be shown +that if all of the above conditions are satisfied, +then the SDE has a unique solution, +which is $\mathcal{F}_t$-adapted, continuous, and exists for all times. + ## References diff --git a/content/know/concept/ito-integral/index.pdc b/content/know/concept/ito-integral/index.pdc index ec49189..cbd4a91 100644 --- a/content/know/concept/ito-integral/index.pdc +++ b/content/know/concept/ito-integral/index.pdc @@ -13,9 +13,8 @@ markup: pandoc # Itō integral The **Itō integral** offers a way to integrate -a time-indexed [random variable](/know/concept/random-variable/) -$G_t$ (i.e. a stochastic process) with respect -to a [Wiener process](/know/concept/wiener-process/) $B_t$, +a given [stochastic process](/know/concept/stochastic-process/) $G_t$ +with respect to a [Wiener process](/know/concept/wiener-process/) $B_t$, which is also a stochastic process. The Itō integral $I_t$ of $G_t$ is defined as follows: @@ -29,7 +28,7 @@ $$\begin{aligned} Where have partitioned the time interval $[a, b]$ into steps of size $h$. The above integral exists if $G_t$ and $B_t$ are adapted -to a common [filtration](/know/concept/sigma-algebra) $\mathcal{F}_t$, +to a common filtration $\mathcal{F}_t$, and $\mathbf{E}[G_t^2]$ is integrable for $t \in [a, b]$. If $I_t$ exists, $G_t$ is said to be **Itō-integrable** with respect to $B_t$. diff --git a/content/know/concept/martingale/index.pdc b/content/know/concept/martingale/index.pdc index 07ed1a4..21fa918 100644 --- a/content/know/concept/martingale/index.pdc +++ b/content/know/concept/martingale/index.pdc @@ -12,15 +12,14 @@ markup: pandoc # Martingale -A **martingale** is a type of stochastic process -(i.e. a time-indexed [random variable](/know/concept/random-variable/)) +A **martingale** is a type of +[stochastic process](/know/concept/stochastic-process/) with important and useful properties, especially for stochastic calculus. For a stochastic process $\{ M_t : t \ge 0 \}$ -on a probability space $(\Omega, \mathcal{F}, P)$ with filtration $\{ \mathcal{F}_t \}$ -(see [$\sigma$-algebra](/know/concept/sigma-algebra/)), -then $\{ M_t \}$ is a martingale if it satisfies all of the following: +on a probability filtered space $(\Omega, \mathcal{F}, \{ \mathcal{F}_t \}, P)$, +then $M_t$ is a martingale if it satisfies all of the following: 1. $M_t$ is $\mathcal{F}_t$-adapted, meaning the filtration $\mathcal{F}_t$ contains enough information @@ -33,19 +32,18 @@ then $\{ M_t \}$ is a martingale if it satisfies all of the following: to be zero $\mathbf{E}(M_t \!-\! M_s | \mathcal{F}_s) = 0$. The last condition is called the **martingale property**, -and essentially means that a martingale is an unbiased random walk. -Accordingly, the [Wiener process](/know/concept/wiener-process/) $\{ B_t \}$ -(Brownian motion) is a prime example of a martingale -(with respect to its own filtration), +and basically means that a martingale is an unbiased random walk. +Accordingly, the [Wiener process](/know/concept/wiener-process/) $B_t$ +(Brownian motion) is an example of a martingale, since each of its increments $B_t \!-\! B_s$ has mean $0$ by definition. Modifying property (3) leads to two common generalizations. -The stochastic process $\{ M_t \}$ above is a **submartingale** +The stochastic process $M_t$ above is a **submartingale** if the current value is a lower bound for the expectation: 3. For $0 \le s \le t$, the conditional expectation $\mathbf{E}(M_t | \mathcal{F}_s) \ge M_s$. -Analogouly, $\{ M_t \}$ is a **supermartingale** +Analogouly, $M_t$ is a **supermartingale** if the current value is an upper bound instead: 3. For $0 \le s \le t$, the conditional expectation $\mathbf{E}(M_t | \mathcal{F}_s) \le M_s$. diff --git a/content/know/concept/random-variable/index.pdc b/content/know/concept/random-variable/index.pdc index 2a8643e..bc41744 100644 --- a/content/know/concept/random-variable/index.pdc +++ b/content/know/concept/random-variable/index.pdc @@ -73,7 +73,8 @@ $$\begin{aligned} \quad \mathrm{for\:any\:} B \in \mathcal{B}(\mathbb{R}^n) \end{aligned}$$ -In other words, for a given Borel set (see $\sigma$-algebra) $B \in \mathcal{B}(\mathbb{R}^n)$, +In other words, for a given Borel set +(see [$\sigma$-algebra](/know/concept/sigma-algebra/)) $B \in \mathcal{B}(\mathbb{R}^n)$, the set of all outcomes $\omega \in \Omega$ that satisfy $X(\omega) \in B$ must form a valid event; this set must be in $\mathcal{F}$. The point is that we need to be able to assign probabilities @@ -94,7 +95,38 @@ $X^{-1}$ can be regarded as the inverse of $X$: it maps $B$ to the event for which $X \in B$. With this, our earlier requirement that $X$ be measurable can be written as: $X^{-1}(B) \in \mathcal{F}$ for any $B \in \mathcal{B}(\mathbb{R}^n)$. -This is also often stated as *"$X$ is $\mathcal{F}$-measurable"*. +This is also often stated as "$X$ is *$\mathcal{F}$-measurable"*. + +Related to $\mathcal{F}$ is the **information** +obtained by observing a random variable $X$. +Let $\sigma(X)$ be the information generated by observing $X$, +i.e. the events whose occurrence can be deduced from the value of $X$, +or, more formally: + +$$\begin{aligned} + \sigma(X) + = X^{-1}(\mathcal{B}(\mathbb{R}^n)) + = \{ A \in \mathcal{F} : A = X^{-1}(B) \mathrm{\:for\:some\:} B \in \mathcal{B}(\mathbb{R}^n) \} +\end{aligned}$$ + +In other words, if the realized value of $X$ is +found to be in a certain Borel set $B \in \mathcal{B}(\mathbb{R}^n)$, +then the preimage $X^{-1}(B)$ (i.e. the event yielding this $B$) +is known to have occurred. + +In general, given any $\sigma$-algebra $\mathcal{H}$, +a variable $Y$ is said to be *"$\mathcal{H}$-measurable"* +if $\sigma(Y) \subseteq \mathcal{H}$, +so that $\mathcal{H}$ contains at least +all information extractable from $Y$. + +Note that $\mathcal{H}$ can be generated by another random variable $X$, +i.e. $\mathcal{H} = \sigma(X)$. +In that case, the **Doob-Dynkin lemma** states +that $Y$ is only $\sigma(X)$-measurable +if $Y$ can always be computed from $X$, +i.e. there exists a function $f$ such that +$Y(\omega) = f(X(\omega))$ for all $\omega \in \Omega$. Now, we are ready to define some familiar concepts from probability theory. The **cumulative distribution function** $F_X(x)$ is @@ -163,6 +195,10 @@ $$\begin{aligned} = \mathbf{E}[X^2] - \big(\mathbf{E}[X]\big)^2 \end{aligned}$$ +It is also possible to calculate expectation values and variances +adjusted to some given event information: +see [conditional expectation](/know/concept/conditional-expectation/). + ## References diff --git a/content/know/concept/sigma-algebra/index.pdc b/content/know/concept/sigma-algebra/index.pdc index 96240ff..94e7306 100644 --- a/content/know/concept/sigma-algebra/index.pdc +++ b/content/know/concept/sigma-algebra/index.pdc @@ -42,9 +42,6 @@ Likewise, a **sub-$\sigma$-algebra** is a sub-family of a certain $\mathcal{F}$, which is a valid $\sigma$-algebra in its own right. - -## Notable applications - A notable $\sigma$-algebra is the **Borel algebra** $\mathcal{B}(\Omega)$, which is defined when $\Omega$ is a metric space, such as the real numbers $\mathbb{R}$. @@ -54,64 +51,6 @@ and all the subsets of $\mathbb{R}$ obtained by countable sequences of unions and intersections of those intervals. The elements of $\mathcal{B}$ are **Borel sets**. -
- -Another example of a $\sigma$-algebra is the **information** -obtained by observing a [random variable](/know/concept/random-variable/) $X$. -Let $\sigma(X)$ be the information generated by observing $X$, -i.e. the events whose occurrence can be deduced from the value of $X$: - -$$\begin{aligned} - \sigma(X) - = X^{-1}(\mathcal{B}(\mathbb{R}^n)) - = \{ A \in \mathcal{F} : A = X^{-1}(B) \mathrm{\:for\:some\:} B \in \mathcal{B}(\mathbb{R}^n) \} -\end{aligned}$$ - -In other words, if the realized value of $X$ is -found to be in a certain Borel set $B \in \mathcal{B}(\mathbb{R}^n)$, -then the preimage $X^{-1}(B)$ (i.e. the event yielding this $B$) -is known to have occurred. - -Given a $\sigma$-algebra $\mathcal{H}$, -a random variable $Y$ is said to be *"$\mathcal{H}$-measurable"* -if $\sigma(Y) \subseteq \mathcal{H}$, -meaning that $\mathcal{H}$ contains at least -all information extractable from $Y$. - -Note that $\mathcal{H}$ can be generated by another random variable $X$, -i.e. $\mathcal{H} = \sigma(X)$. -In that case, the **Doob-Dynkin lemma** states -that $Y$ is only $\sigma(X)$-measurable -if $Y$ can always be computed from $X$, -i.e. there exists a function $f$ such that -$Y(\omega) = f(X(\omega))$ for all $\omega \in \Omega$. - -
- -The concept of information can be extended for -stochastic processes (i.e. time-indexed random variables): -if $\{ X_t : t \ge 0 \}$ is a stochastic process, -its **filtration** $\mathcal{F}_t$ contains all -the information generated by $X_t$ up to the current time $t$: - -$$\begin{aligned} - \mathcal{F}_t - = \sigma(X_s : 0 \le s \le t) -\end{aligned}$$ - -In other words, $\mathcal{F}_t$ is the "accumulated" $\sigma$-algebra -of all information extractable from $X_t$, -and hence grows with time: $\mathcal{F}_s \subset \mathcal{F}_t$ for $s < t$. -Given $\mathcal{F}_t$, all values $X_s$ for $s \le t$ can be computed, -i.e. if you know $\mathcal{F}_t$, then the present and past of $X_t$ can be reconstructed. - -Given some filtration $\mathcal{H}_t$, a stochastic process $X_t$ -is said to be *"$\mathcal{H}_t$-adapted"* -if $X_t$'s own filtration $\sigma(X_s : 0 \le s \le t) \subseteq \mathcal{H}_t$, -meaning $\mathcal{H}_t$ contains enough information -to determine the current and past values of $X_t$. -Clearly, $X_t$ is always adapted to its own filtration. - ## References diff --git a/content/know/concept/stochastic-process/index.pdc b/content/know/concept/stochastic-process/index.pdc new file mode 100644 index 0000000..5d50da8 --- /dev/null +++ b/content/know/concept/stochastic-process/index.pdc @@ -0,0 +1,62 @@ +--- +title: "Stochastic process" +firstLetter: "S" +publishDate: 2021-11-07 +categories: +- Mathematics + +date: 2021-11-07T18:45:42+01:00 +draft: false +markup: pandoc +--- + +# Stochastic process + +A **stochastic process** $X_t$ is a time-indexed +[random variable](/know/concept/random-variable/), +$\{ X_t : t > 0 \}$, i.e. a set of (usually correlated) +random variables, each labelled with a unique timestamp $t$. + +Whereas "ordinary" random variables are defined on +a probability space $(\Omega, \mathcal{F}, P)$, +stochastic process are defined on +a **filtered probability space** $(\Omega, \mathcal{F}, \{ \mathcal{F}_t \}, P)$. +As before, $\Omega$ is the sample space, +$\mathcal{F}$ is the event space, +and $P$ is the probability measure. + +The **filtration** $\{ \mathcal{F}_t : t \ge 0 \}$ +is a time-indexed set of [$\sigma$-algebras](/know/concept/sigma-algebra/) on $\Omega$, +which contains at least all the information generated +by $X_t$ up to the current time $t$, +and is a subset of $\mathcal{F}_t$: + +$$\begin{aligned} + \mathcal{F} + \supseteq \mathcal{F}_t + \supseteq \sigma(X_s : 0 \le s \le t) +\end{aligned}$$ + +In other words, $\mathcal{F}_t$ is the "accumulated" $\sigma$-algebra +of all information extractable from $X_t$, +and hence grows with time: $\mathcal{F}_s \subseteq \mathcal{F}_t$ for $s < t$. +Given $\mathcal{F}_t$, all values $X_s$ for $s \le t$ can be computed, +i.e. if you know $\mathcal{F}_t$, then the present and past of $X_t$ can be reconstructed. + +Given any filtration $\mathcal{H}_t$, a stochastic process $X_t$ +is said to be *"$\mathcal{H}_t$-adapted"* +if $X_t$'s own filtration $\sigma(X_s : 0 \le s \le t) \subseteq \mathcal{H}_t$, +meaning $\mathcal{H}_t$ contains enough information +to determine the current and past values of $X_t$. +Clearly, $X_t$ is always adapted to its own filtration. + +Filtration and their adaptations are very useful +for working with stochastic processes, +most notably for calculating [conditional expectations](/know/concept/conditional-expectation/). + + + +## References +1. U.H. Thygesen, + *Lecture notes on diffusions and stochastic differential equations*, + 2021, Polyteknisk Kompendie. diff --git a/content/know/concept/wiener-process/index.pdc b/content/know/concept/wiener-process/index.pdc index 3602b44..f8610a2 100644 --- a/content/know/concept/wiener-process/index.pdc +++ b/content/know/concept/wiener-process/index.pdc @@ -13,14 +13,13 @@ markup: pandoc # Wiener process -The **Wiener process** is a stochastic process that provides -a pure mathematical definition of the physical phenomenon of **Brownian motion**, +The **Wiener process** is a [stochastic process](/know/concept/stochastic-process/) +that provides a pure mathematical definition +of the physical phenomenon of **Brownian motion**, and hence is also called *Brownian motion*. A Wiener process $B_t$ is defined as any -time-indexed [random variable](/know/concept/random-variable/) -$\{B_t: t \ge 0\}$ (i.e. stochastic process) -that has the following properties: +stochastic process $\{B_t: t \ge 0\}$ that satisfies: 1. Initial condition $B_0 = 0$. 2. Each **increment** of $B_t$ is independent of the past: @@ -49,28 +48,7 @@ Another consequence is invariance under "time inversion", by defining $\sqrt{\alpha} = t$, such that $W_t = t B_{1/t}$. Despite being continuous by definition, -the **total variation** $V(B)$ of $B_t$ is infinite -(informally, the curve is infinitely long). -For $t_i \in [0, 1]$ in $n$ steps of maximum size $\Delta t$: - -$$\begin{aligned} - V_t - = \lim_{\Delta t \to 0} \sup \sum_{i = 1}^n \big|B_{t_i} - B_{t_{i-1}}\big| - = \infty -\end{aligned}$$ - -However, curiously, the **quadratic variation**, written as $[B]_t$, -turns out to be deterministically finite and equal to $t$, -while a differentiable function $f$ would have $[f]_t = 0$: - -$$\begin{aligned} - \:[B]_t - = \lim_{\Delta t \to 0} \sum_{i = 1}^n \big|B_{t_i} - B_{t_{i - 1}}\big|^2 - = t -\end{aligned}$$ - -Therefore, despite being continuous by definition, -the Wiener process is not differentiable, +the Wiener process is not differentiable in general, not even in the mean square, because: $$\begin{aligned} diff --git a/content/know/concept/young-dupre-relation/index.pdc b/content/know/concept/young-dupre-relation/index.pdc index d3f36cb..579bd5e 100644 --- a/content/know/concept/young-dupre-relation/index.pdc +++ b/content/know/concept/young-dupre-relation/index.pdc @@ -81,7 +81,7 @@ $$\begin{aligned} = \alpha_{sg} - \alpha_{sl} \end{aligned}$$ -At the edge of the droplet, imagine a small rectangular triangle +At the edge of the droplet, imagine a small right-angled triangle with one side $\dd{x}$ on the $x$-axis, the hypotenuse on $y(x)$ having length $\dd{x} \sqrt{1 + (y')^2}$, and the corner between them being the contact point with angle $\theta$. -- cgit v1.2.3