-
Notifications
You must be signed in to change notification settings - Fork 0
/
whatplaysdice.tex
242 lines (174 loc) · 7.97 KB
/
whatplaysdice.tex
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
\documentclass[11pt]{article}
\usepackage{amsthm}
\usepackage{mathtools}
\usepackage[inline]{enumitem}
\usepackage{ifthen}
\usepackage[utf8]{inputenc} %allows non-ascii in bib file
\usepackage{xcolor}
\usepackage[backend=biber, backref=true, maxbibnames = 10, style = alphabetic]{biblatex}
\usepackage[bookmarks=true, colorlinks=true, linkcolor=blue!50!black,
citecolor=orange!50!black, urlcolor=orange!50!black, pdfencoding=unicode]{hyperref}
\usepackage[capitalize]{cleveref}
\usepackage{tikz}
\usepackage{amssymb}
\usepackage{mathrsfs}
\usepackage{dutchcal}
\usepackage{fontawesome}
\usepackage{stmaryrd}
% cleveref %
\newcommand{\creflastconjunction}{, and\nobreakspace} % serial comma
\crefformat{enumi}{\card#2#1#3}
\crefalias{chapter}{section}
% hyperref %
\hypersetup{final}
% enumitem %
\setlist{nosep}
\setlistdepth{6}
% tikz %
\usetikzlibrary{
cd,
math,
decorations.markings,
decorations.pathreplacing,
positioning,
arrows.meta,
shapes,
shadows,
shadings,
calc,
fit,
quotes,
intersections,
circuits,
circuits.ee.IEC
}
%-------- Renewed commands --------%
\renewcommand{\ss}{\subseteq}
%-------- Other Macros --------%
\DeclareMathOperator{\Hom}{Hom}
\DeclareMathOperator{\Mor}{Mor}
\DeclareMathOperator{\dom}{dom}
\DeclareMathOperator{\cod}{cod}
\DeclareMathOperator{\idy}{idy}
\DeclareMathOperator{\comp}{com}
\DeclareMathOperator*{\colim}{colim}
\DeclareMathOperator{\im}{im}
\DeclareMathOperator{\ob}{Ob}
\DeclareMathOperator{\Tr}{Tr}
\DeclareMathOperator{\el}{El}
\DeclareMathOperator{\Funn}{Fun}
\newcommand{\Set}[1]{\mathsf{#1}}%a named set
\newcommand{\ord}[1]{\mathsf{#1}}%an ordinal
\newcommand{\cat}[1]{\mathcal{#1}}%a generic category
\newcommand{\Cat}[1]{\mathbf{#1}}%a named category
\newcommand{\fun}[1]{\mathrm{#1}}%a function
\newcommand{\Fun}[1]{\mathit{#1}}%a named functor
\newcommand{\N}{\mathbb{N}}
\newcommand{\ox}{\otimes}
\newcommand{\Poly}{{\sf Poly}}
\newcommand{\id}{\mathrm{id}}
\newcommand{\then}{\mathbin{\fatsemi}}
\newcommand{\cocolon}{:\!}
\newcommand{\iso}{\cong}
\newcommand{\too}{\longrightarrow}
\newcommand{\tto}{\rightrightarrows}
\newcommand{\To}[2][]{\xrightarrow[#1]{#2}}
\renewcommand{\Mapsto}[1]{\xmapsto{#1}}
\newcommand{\Tto}[3][13pt]{\begin{tikzcd}[sep=#1, cramped, ampersand replacement=\&, text height=1ex, text depth=.3ex]\ar[r, shift left=2pt, "#2"]\ar[r, shift right=2pt, "#3"']\&{}\end{tikzcd}}
\newcommand{\Too}[1]{\xrightarrow{\;\;#1\;\;}}
\newcommand{\from}{\leftarrow}
\newcommand{\fromm}{\longleftarrow}
\newcommand{\ffrom}{\leftleftarrows}
\newcommand{\From}[1]{\xleftarrow{#1}}
\newcommand{\Fromm}[1]{\xleftarrow{\;\;#1\;\;}}
\newcommand{\surj}{\twoheadrightarrow}
\newcommand{\inj}{\hookrightarrow}
\newcommand{\wavyto}{\rightsquigarrow}
\newcommand{\lollipop}{\multimap}
\newcommand{\imp}{\Rightarrow}
\newcommand{\down}{\mathbin{\downarrow}}
\newcommand{\fromto}{\leftrightarrows}
\newcommand{\tickar}{\xtickar{}}
\newcommand{\slashar}{\xslashar{}}
\newcommand{\inv}{^{-1}}
\newcommand{\op}{^\tn{op}}
\newcommand{\tn}[1]{\textnormal{#1}}
\newcommand{\ol}[1]{\overline{#1}}
\newcommand{\ul}[1]{\underline{#1}}
\newcommand{\wt}[1]{\widetilde{#1}}
\newcommand{\wh}[1]{\widehat{#1}}
\newcommand{\wc}[1]{\widecheck{#1}}
\newcommand{\ubar}[1]{\underaccent{\bar}{#1}}
\newcommand{\bb}{\mathbb{B}}
\newcommand{\cc}{\mathbb{C}}
\newcommand{\nn}{\mathbb{N}}
\newcommand{\pp}{\mathbb{P}}
\newcommand{\qq}{\mathbb{Q}}
\newcommand{\zz}{\mathbb{Z}}
\newcommand{\rr}{\mathbb{R}}
\newcommand{\finset}{\Cat{Fin}}
\newcommand{\smset}{\Cat{Set}}
\newcommand{\lgset}{\Cat{SET}}
\newcommand{\smcat}{\Cat{Cat}}
\newcommand{\lgcat}{\Cat{CAT}}
\newcommand{\prof}{\mathbb{P}\Cat{rof}}
\newcommand{\set}{\tn{-}\Cat{Set}}
\newcommand{\coalg}{\tn{-}\Cat{Coalg}}
\newcommand{\const}[1][\blank]{#1 !}
\newcommand{\hh}[2][]{#1 \tn{\textit{#2}} #1}
\newcommand{\qqand}{\hh[\qquad]{and}}
\newcommand{\qand}{\hh[\quad]{and}}
\renewcommand{\iff}[1][\;\;]{#1\Leftrightarrow#1}
\newcommand{\ifff}[1][\;\;]{#1\xLeftrightarrow{\quad}#1}
\newcommand{\hi}[4][]{#1 #2 \tn{\textit{#4}} #3}
\newcommand{\where}[1][,]{\hi[#1]{\qquad}{\quad}{where}}
\newcommand{\qimplies}{\hh[\quad]{$\implies$}}
\newcommand{\lott}{\mathit{lott}}
\newcommand{\yon}{\mathcal{y}}
\newcommand{\dnote}[1]{{\quad \color{blue}$\lozenge$\;David says:}~#1\;{\color{blue}$\lozenge$}\quad}
\newcommand{\pnote}[1]{{\quad \color{red}$\lozenge$\;Priyaa says:}~#1\;{\color{red}$\lozenge$}\quad}
\title{The ``typical'' $[\lott,\yon]$-coalgebra}
\title{Does God play dice? Or...\\What's the best $[\lott,\yon]$-coalgebra?}
\author{
David I.\ Spivak \quad \quad \quad Priyaa Varshinee Srinivasan}
\date{\vspace{-.1in}}
\begin{document}
\maketitle
\begin{abstract}
\end{abstract}
\section{Introduction}
Einstein famously said ``God does not play dice" in response to... quantum something. But also in computers we need to \emph{sample} from probability distributions. So what does play dice? In this post we'll explain how this question relates to another one: what is the best $[\lott,\yon]$-coalgebra? We'll explain how a $[\lott,\yon]$-coalgebra is a winning-ticket-picker for lotteries, and explain a few senses in which one may consider one the best!
\paragraph{Acknowledgements.} Thanks to Maxine Collard for useful conversations.
\section{Reviewing polynomials and lotteries}
See other \href{https://topos.site/blog/2023-03-23-distributions-and-lotteries/}{blog post}, pick interesting things. For example, the polynomial, what the monad multiplication means. Also, recall $\otimes$ and $[-,-]$ and what $[\lott,\yon]$ means.
%
%For any finite set $N$, define $\Delta_N$ to be the set of probability distributions on $N$,
%
%\[ \Delta_N := \left \{ P: N \to [0,1] \mid 1 = \sum_{n:N} P_n \right \} \]
%
%\[ \lott = \sum_{N:\N} \sum_{p: \Delta_N} y^N \]
%
%For any two polynomials $p = \sum_{P:p(1)}$ and $q$,
%\begin{align*}
% p \ox q &= \sum_{P: p(1)} \yon^{p[P]} \ox \sum_{Q: q(1)} \yon^{q[Q]} := \sum_{(P,Q): p(1) \times q(1)} \yon^{p[P] \times q[Q]} \\
% [p,q] &= \prod_{P:p(1)} \sum_{Q: q(1)} \prod_{e: q[Q]} \sum_{d:p[P]} 1
% \end{align*}
%
%\begin{align}
%\nonumber
%\Poly(p,q) &= \Poly\left(\sum_{i:P} \yon^{p[P]}, q\right) \\\nonumber
%&\cong \prod_{P:p(1)} \Poly\left(\yon^{p[P]}, q\right) & \text{Universal property of coproduct}\\ \nonumber
%&\cong \prod_{P:p(1)} q(p[P]) & \text{Yoneda Lemma} \\ \nonumber
%& \cong \prod_{P:p(1)} \sum_{Q: q(1)} p[P]^{q[Q]} \\\label{eqn.poly_map}
%& = \prod_{P:p(1)} \sum_{Q: q(1)} \prod_{e: q[Q]} \sum_{d:p[P]} 1
%\end{align}
\section{Typical sequences and Martin-L\"of randomness}
\dnote{Priyaa, see \url{https://chatgpt.com/share/20a5704c-6c87-4f49-9b57-4e632c312e38}}
A $[\lott,\yon]$-coalgebra consists of a set $S$ and a function $S\to[\lott,\yon](S)$. We say that it is \emph{typical} if, for any $s\in S$ the sequence is typical. We say that it is \emph{algorithmically random} if, for any $s\in S$ the sequence is algorithmically random.
\section{The ``God machine''}
Albert Einstein's God did not play dice, but perhaps David Bohm's ``guiding wave'' does. Here we consider the idea of a machine, which Priyaa calls the "God machine", that picks a winning ticket from any named lottery.
Construction: for each lottery $(N,P)$, a stream $s_{N,P}:\nn\to N$ that's typical, algorithmically random, etc. The coalgebra outputs these, and then for each actual choice $(N,P)$ of lottery played, it rips off one ticket. (Write this as a formula.)
This has the property that for any list $(N_1,P_1),\ldots(N_k,P_k)$ with each $(N_i,P_i)\neq (N,P)$, one has $(s.(N_1,P_1)\ldots(N_k,P_k))(N,P))=s(N,P)$, i.e.\ that playing one lottery cannot affect which ticket is poised to win another lottery.
Prove that if $s$ is typical, then for any lottery $(N,P)$, the stream $s.(N,P)$ is also typical; same for algorithmically random. This means that typicality and algorithmic randomness define propositions in the internal logic of the topos $[\lott,\yon]\coalg$.
That is, you might say that the terminal coalgebra such that every element of it is typical is best!
\end{document}