From 87c1a759e9db2e0939dcf4c5a3ba7e9be0e7f86b Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Th=C3=A9o=20Winterhalter?= Date: Fri, 29 May 2020 17:11:15 +0200 Subject: [PATCH] Pass on chap 8 (Matthieu and I) --- chapters/models.tex | 45 +++++++++++++++++++++++++-------------------- main.bib | 17 +++++++++++++++++ 2 files changed, 42 insertions(+), 20 deletions(-) diff --git a/chapters/models.tex b/chapters/models.tex index 21a8367..29f5940 100644 --- a/chapters/models.tex +++ b/chapters/models.tex @@ -32,6 +32,8 @@ \subsection{What can be proved using models} \emph{know} that a theory is consistent, instead we should see it as a theory we \emph{trust}, typically a theory that is widely accepted as beeing consistent by the community of mathematicians. +In this case we talk about \emph{relative consistency}: the theory is +consistent, relative to the consistency of the theory of its model. This also applies to the meta-theory in which we show that the interpretation is correct. As such, it is best to keep it as simple as possible to avoid relying on the consistency of too complicated objects. @@ -52,7 +54,7 @@ \subsection{What can be proved using models} could be proven in \cT, then it would be valid in both models, leading to at least one of them being inconsistent. -The fact that a proposition can neither be validated or invalidated in a theory +The fact that a proposition can neither be validated nor invalidated in a theory can come as surprising for some, especially in a classical mindset. \reminder[-1.8cm]{Classical logic}{ Classical logic is often characterised by the presence of the \acrshort{LEM} @@ -91,7 +93,7 @@ \section{Set-theoretic models} are interpreted as the singleton \(\{ \emptyset \}\). First, I will remind what rules we are dealing with. This theory will only have -two universes: \(\Prop\) and \(\Type\), the model can be adapted to deal with +two universes: \(\Prop\) and \(\Type\). The model can be adapted to deal with more universes but it is a bit more complicated. \marginnote[1.7cm]{ We have two rules for \(\Pi\)-types because we have an impredicative @@ -207,16 +209,18 @@ \section{Set-theoretic models} \[ \Ga = x_n : A_n, \dots x_i : A_i, \dots x_0 : A_0 \] -I do not really believe you can distinguish between propositions and usual types -in the current setting so in~\sidecite{aczel1998relating}, Aczel introduces a -non-standard encoding of set-theoretic functions such that the constant function +One cannot distinguish between propositions and usual types in the current +setting so in~\sidecite{aczel1998relating}, Aczel introduces a non-standard +encoding of set-theoretic functions such that the constant function \(x \mapsto 0\) is \emph{equal} to \(0\). +In \sidecite{timany:hal-01615123} they build a set-theoretic model for the whole +of \acrshort{PCUIC} using the same trick. I did not precise what \(\mathcal{U}\) was. From \(\mathcal{U}\) we only require -that it is closed under product and power-sets and contains at least \(2\). +that it is closed under products and power-sets and contains at least \(2\). It needs only be the cardinal \(\aleph_0\) in our case, but if we wanted more universes---say a hierarchy of them---we could use inaccessible cardinals. -We discuss see them briefly in \refsec{cat-models}. +We discuss them briefly in \refsec{cat-models}. \paradot{Consistency} @@ -263,7 +267,7 @@ \section{Set-theoretic models} \transl{\Ga \vdash \lambda (x:A). t}(\gamma) \in \transl{\Ga \vdash \Pi (x:A). B}(\gamma) \] -If we reformulate we want to show that for all \(\gamma \in \Ga\) and +If we reformulate, we want to show that for all \(\gamma \in \Ga\) and \(a \in \transl{\Ga \vdash A}(\gamma)\) we have \[ \transl{\Ga, x : A \vdash t}(\gamma, a) \in @@ -282,7 +286,7 @@ \section{Set-theoretic models} computational content since \(\beta\)-convertible terms are interpreted as the same set. They provide however a nice way of seeing the relation between set-theory---the widely accepted framework in which to do mathematics---and type -theory which we\sidenote{I'll leave that one as imprecise as it is.} advocate +theory which we\sidenote{I will leave that one as imprecise as it is.} advocate instead. \section{Categorical models} @@ -485,10 +489,10 @@ \subsection{Categories with families} \end{itemize} \marginnote[-1.3cm]{ For the successor I use \(\Pi\)-types when I could have instead put things - in the context. Both are valid and I'm often favorable to the idea that things - should be kept separate as much as possible, but the reliance on application - and everything allows us to avoid proving the same things several times, - especially since manipulating contexts is so tedious with categories. + in the context. Both are valid and I am often favorable to the idea that + things should be kept separate as much as possible, but the reliance on + application and everything allows us to avoid proving the same things several + times, especially since manipulating contexts is so tedious with categories. } such that \[ @@ -808,7 +812,7 @@ \subsection{Cardinal model} {\eqterm{\Ga}{u}{v}{A}} % \end{equation*} - See \refdef{reflection} + See \refdef{reflection}. }% % \begin{proof} @@ -836,7 +840,7 @@ \subsection{Cardinal model} \end{definition} This consists in a strong invariance principle stating that isomorphic types -cannot be distinguished. Another invariance principle is the univalence axiom. +cannot be distinguished. Univalence is another invariance principle. \begin{theorem} The cardinal model validates isomorphism reflection. @@ -1048,8 +1052,9 @@ \subsection{Cardinal model} where \(\aleph_\mu = \prod_{x \in \CEl{a(\gamma)}} \CEl{\beta^{-1}(b(\gamma))(x)}\). -I will not really go into detail about cardinals and everything here, or as to -why this universe also supports natural numbers, equality and such. +I will stop here regarding cardinals, or as to why this universe also supports +natural numbers, equality and such, because the details do not provide new +insights about this model. \paradot{A warning about syntax} @@ -1107,15 +1112,15 @@ \subsection{Cardinal model} \[ \beta^{-1}(\beta(x \mapsto x))(2) = (x \mapsto x)(2) \] -where simplify the expression \(\beta^{-1} \circ \beta\) to the identity -function. As usual I left the subscripts implicit, but that is where the error +where we simplified the expression \(\beta^{-1} \circ \beta\) to the identity +function. As usual, I left the subscripts implicit, but that is where the error is coming from: we actually have \[ \beta_{2^\aleph_0}^{-1} \circ \beta_{{\aleph_0}^{\aleph_0}} \] which do not cancel each other out. -This is case for using type annotations on \(\lambda\)-abstractions and +This is a case for the use type annotations on \(\lambda\)-abstractions and applications to block \(\beta\)-reduction when the types do not match. \section{Type-theoretic models} diff --git a/main.bib b/main.bib index 93bd740..9031fc8 100644 --- a/main.bib +++ b/main.bib @@ -1024,4 +1024,21 @@ @phdthesis{haselwarterEffective author={Haselwarter, Philipp G.}, year={2020}, school={University of Ljubljana} +} + +@techreport{timany:hal-01615123, + TITLE = {{Consistency of the Predicative Calculus of Cumulative Inductive Constructions (pCuIC)}}, + AUTHOR = {Timany, Amin and Sozeau, Matthieu}, + URL = {https://hal.inria.fr/hal-01615123}, + NOTE = {Version 2 fixes some typos from version 1.Version 3 fixes a typo in a typing rule from version 2.}, + TYPE = {Research Report}, + NUMBER = {RR-9105}, + PAGES = {30}, + INSTITUTION = {{KU Leuven, Belgium ; Inria Paris}}, + YEAR = {2017}, + MONTH = Oct, + KEYWORDS = {logic ; metatheory ; calculus of constructions ; set theory ; calcul des constructions ; Coq ; m{\'e}tath{\'e}orie ; th{\'e}orie des ensembles ; logique}, + PDF = {https://hal.inria.fr/hal-01615123/file/consistency.pdf}, + HAL_ID = {hal-01615123}, + HAL_VERSION = {v3}, } \ No newline at end of file