\( \newcommand{\matr}[1] {\mathbf{#1}} \newcommand{\vertbar} {\rule[-1ex]{0.5pt}{2.5ex}} \newcommand{\horzbar} {\rule[.5ex]{2.5ex}{0.5pt}} \newcommand{\E} {\mathrm{E}} \)
deepdream of a sidewalk
Show Answer
\( \newcommand{\cat}[1] {\mathrm{#1}} \newcommand{\catobj}[1] {\operatorname{Obj}(\mathrm{#1})} \newcommand{\cathom}[1] {\operatorname{Hom}_{\cat{#1}}} \newcommand{\multiBetaReduction}[0] {\twoheadrightarrow_{\beta}} \newcommand{\betaReduction}[0] {\rightarrow_{\beta}} \newcommand{\betaEq}[0] {=_{\beta}} \newcommand{\string}[1] {\texttt{"}\mathtt{#1}\texttt{"}} \newcommand{\symbolq}[1] {\texttt{`}\mathtt{#1}\texttt{'}} \newcommand{\groupMul}[1] { \cdot_{\small{#1}}} \newcommand{\groupAdd}[1] { +_{\small{#1}}} \newcommand{\inv}[1] {#1^{-1} } \newcommand{\bm}[1] { \boldsymbol{#1} } \require{physics} \require{ams} \require{mathtools} \)
Math and science::INF ML AI

Monte Carlo Methods

The this card describes Monte Carlo methods in the words of David MacKay. A lot of this card's content is drawn from Chapter 29 of MacKay's book.

Monte Carlo methods. Definition.

Monte Carlo methods is a term for something that utilizes a random number generator to solve one or both of the following problems.

  1. Generate [something] from a given [what?]
  2. Estimate expectations of functions under this probability distribution. For example, for some function \( \phi \), we want to know:
    [\[\int {\dd}^N \vb{x} \;\; ? \;\; \phi(\vb{x}) \]]

The notation used above does a lot of heavy lifting. "\( \mathbb{P}(\vb{x}) \)" suggests a set \( L \subset \mathbb{R}^N \) is the co-domain of an implied random variable \( Y : ? \to L \). \( L \) is also the domain of the function \( \mathbb{P} \) which must have a signature of the form \( \mathbb{P} : L \to \mathbb{R} \). The function \( \phi \) seems to have a signature \( \phi : L \to \mathbb{R} \). It is composed with \( Y \) like \( \phi \circ Y \) and thus, \( \phi \) is a random variable. The composition implies that the domain of \( \phi \) is \( L \). The integration formula further implies either that the co-domain of \( \phi \) must be \( \mathbb{R} \) (alternatively, the integration is syntax is allowing for the interpretation of a vector valued integration which seems a stretch).

The reverse side describes Lake plankton analogy to the above problems. Can you remember the details? The reverse side also rewords the above description in the language of probability distributions.