Probability integral as a linearization
Abstract
In fuzzified probability theory, a classical probability space
$(\Omega,\mathbf{A},p)$ is replaced by a generalized probability space
$(\Omega, \mathcal{M}(\mathbf{A}), \int\! (.) \textup{d}p)$, where
$\mathcal{M}(\mathbf{A})$ is the set of all measurable functions into [0,1]
and $\int\! (.) \textup{d}p $ is the probability integral with respect to $p$.
Our paper is devoted to the transition from $p$ to $\int\! (.) \textup{d}p$. The transition
is supported by the following categorical argument: there is a minimal category
and its epireflective subcategory such that $\mathbf{A}$ and $\mathcal{M}(\mathbf{A})$
are objects, probability measures and probability integrals are morphisms,
$\mathcal{M}(\mathbf{A})$ is the epireflection of ~$\mathbf{A}$, $\int\! (.) \textup{d}p$
is the corresponding unique extension of ~$p$, and $\mathcal{M}(\mathbf{A})$
carries the initial structure with respect to probability integrals.
We discuss reasons why the fuzzy random events are modeled by $\mathcal{M}(\mathbf{A})$
equipped with pointwise partial order, pointwise \L ukasiewicz operations
(logic) and pointwise sequential convergence. Each probability measure induces on
classical random events an additive linear preorder which helps making decisions.
We show that probability integrals can be characterized as the additive linearizations
on fuzzy random events, i.e., sequentially continuous maps, preserving order, top and bottom
elements.
$(\Omega,\mathbf{A},p)$ is replaced by a generalized probability space
$(\Omega, \mathcal{M}(\mathbf{A}), \int\! (.) \textup{d}p)$, where
$\mathcal{M}(\mathbf{A})$ is the set of all measurable functions into [0,1]
and $\int\! (.) \textup{d}p $ is the probability integral with respect to $p$.
Our paper is devoted to the transition from $p$ to $\int\! (.) \textup{d}p$. The transition
is supported by the following categorical argument: there is a minimal category
and its epireflective subcategory such that $\mathbf{A}$ and $\mathcal{M}(\mathbf{A})$
are objects, probability measures and probability integrals are morphisms,
$\mathcal{M}(\mathbf{A})$ is the epireflection of ~$\mathbf{A}$, $\int\! (.) \textup{d}p$
is the corresponding unique extension of ~$p$, and $\mathcal{M}(\mathbf{A})$
carries the initial structure with respect to probability integrals.
We discuss reasons why the fuzzy random events are modeled by $\mathcal{M}(\mathbf{A})$
equipped with pointwise partial order, pointwise \L ukasiewicz operations
(logic) and pointwise sequential convergence. Each probability measure induces on
classical random events an additive linear preorder which helps making decisions.
We show that probability integrals can be characterized as the additive linearizations
on fuzzy random events, i.e., sequentially continuous maps, preserving order, top and bottom
elements.
Full Text:
Subscribers OnlyDOI: https://doi.org/10.2478/tmmp-2018-0017