-
Notifications
You must be signed in to change notification settings - Fork 20
/
1-logic.tex
702 lines (642 loc) · 28.7 KB
/
1-logic.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
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
\chapter{Logic}
Before we can formalize \emph{learning} about the world around us we
first need a formal means of \emph{describing} the world. \emph{Logical
propositions} provide a mathematically-precise language for quantifying
information about a system with certainty. In this section we will review
abstract logical propositions, their manipulations, and how we can
represent them in practice. We conclude with a brief discussion of
implicative statements.
\section{Logical Propositions}
If $\mathcal{S}$ is an abstract system of interest then any description of
$\mathcal{S}$ can be defined as a \emph{logical proposition} which asserts
that a statement about that system is either true or false. For example,
consider describing one of the eight planets in our Solar System: Mercury
($\mercury$), Venus ($\venus$), Earth ($\earth$), Mars ($\mars$), Jupiter
($\jupiter$), Saturn ($\saturn$), Uranus ($\uranus$), and Neptune ($\neptune$).
Information about the planet is encapsulated in logical propositions asserting
both statements that are true, for example
%
\begin{equation*}
P_{1} = \text{``the planet has an atmosphere''},
\end{equation*}
%
and statements that are false, for example
\begin{equation*}
P_{2}=\text{``the planet does not have rings''}.
\end{equation*}
Logical propositions can be manipulated into other logical propositions
with any of three \emph{boolean operations} (Table
\ref{tab:boolean_truth_tables}). \emph{Negatation}, $\neg$, is a unary
operator resulting in a proposition asserting the converse of the input
proposition:
%
\begin{equation*}
\neg P_{1}=\text{``the planet does \emph{not} have an atmosphere''}.
\end{equation*}
%
Binary operations take propositions about two statements and define
a new proposition about the joint statement. The result of a
\emph{conjunction}, $\wedge$, is true only when both input propositions
are true:
%
\begin{equation*}
P_{1} \wedge P_{2} =
\text{``the planet does not has both an atmosphere and rings''}.
\end{equation*}
%
A \emph{disjunction}, $\vee$, however, is true whenever either input
proposition is true :
%
\begin{equation*}
P_{1} \vee P_{2} =
\text{``the planet does have an atmosphere \emph{or} rings''}.
\end{equation*}
\begin{table}
\centering
\renewcommand{\arraystretch}{1.5}
\begin{tabular}{cc}
\rowcolor[gray]{0.9}
\multicolumn{2}{c}{\textbf{Negation}} \\
%
\rowcolor[gray]{0.9}
$P$ & $\neg P$ \\
%
False & True \\
True & False \\
\end{tabular}
%
\vspace{3mm} \\
%
\begin{tabular}{ccc}
\rowcolor[gray]{0.9}
\multicolumn{3}{c}{\textbf{Conjunction}} \\
%
\rowcolor[gray]{0.9}
$P_{1}$ & $P_{2}$ & $P_{1} \wedge P_{2}$ \\
%
False & False & False \\
True & False & False \\
False & True & False \\
True & True & True \\
\end{tabular}
%
\hspace{5mm}
%
\begin{tabular}{ccc}
\rowcolor[gray]{0.9}
\multicolumn{3}{c}{\textbf{Disjunction}} \\
%
\rowcolor[gray]{0.9}
$P_{1}$ & $P_{2}$ & $P_{1} \vee P_{2}$ \\
%
False & False & False \\
True & False & True \\
False & True & True \\
True & True & True \\
\end{tabular}
\caption{The three basic Boolean operations transform given
logical propositions into new logical propositions, and their
action is conveniently summarized in truth tables.}
\label{tab:boolean_truth_tables}
\end{table}
Naively, we might think that a system would be completely described only
once we have defined a proposition for every statement, classifying each
as either true or false. In his infamous Incompleteness Theorems, however,
G\"{o}del showed that if we try to define a proposition for every possible
statement about $\mathcal{S}$ then there will exist some pathological
propositions that are not consistent with the Boolean operations. For
example, there are always pathological propositions asserting truth whose
negations do not assert falseness! These pathological propositions cannot
be explicitly constructed in practice, only proven to exist, but in order to
ensure a self-consistent description of the system we need some means of
removing them.
To avoid these pathological propositions we leverage the fact that propositions
about non-pathological statements yield non-pathological propositions when
we apply any Boolean operation. In other words, well-behaved propositions
are \emph{closed} under these operations, and any collection of well-behaved,
consistent propositions forms a logical algebra,
$\mathcal{L} \! \left( \mathcal{S} \right)$.
We will refer to the pairing of an abstract system with some choice of a logical
algebra as a \emph{logical model},
$\left( \mathcal{S}, \mathcal{L} \! \left( \mathcal{S} \right) \right)$.
\section{Representations of Logical Models}
Logical models provide a very abstract, and hence very genetic, framework for
describing arbitrary systems, but that abstraction also makes these descriptions
ungainly to specify and manipulate in practice. For our abstract descriptions to
be useful we need an explicit language in which we can communicate them,
which we will refer to as a \emph{representation}
(Table \ref{tab:representation_examples}).
\begin{table}
\centering
\renewcommand{\arraystretch}{1.5}
\begin{tabular}{ccc}
\rowcolor[gray]{0.9}
\textbf{Abstract System} & \textbf{Representation} & \textbf{Measurable Maps} \\
%
Historical Decrees & Ancient Languages & Translations \\
& (Egyptian, Demotic, Green, $\ldots$) & (Rosetta Stone) \\
%
Computer Program & Programming Languages & Source-to-Source \\
& (C++, S, Python, $\ldots$) & Compilers \\
%
Unordered Categories & Integers & Permutations \\
%
Direction & Vector Space & Changes of Basis \\
%
Position & Coordinates & Reparameterizations \\
%
Quantification & Units & Changes of Units \\
\end{tabular}
\caption{Representations provide a concrete language in which we can
communicate the descriptions of an abstract system. Because there
are many degenerate ways to express these descriptions, these
representations are not unique and in practice we have to be careful
that the irrelevant details of any particular representation do not affect
the descriptions themselves.}
\label{tab:representation_examples}
\end{table}
In order to construct a representation we assume that the properties
of our abstract system take values in some \emph{sample space},
$\Theta$, such as integers or real numbers. Statements about these
properties are represented as subsets of the sample space, and
logical propositions are represented as set inclusion. If a statement
is represented by $E \subset \Theta$, for example, then a logical
proposition asserting that the statement is true is represented by
asserting that $\theta \in E$, and one asserting that the statement
is false is represented by asserting that $\theta \notin E$.
With propositions represented by set inclusion and set exclusion,
the Boolean operations are naturally implemented with set operations.
Conjunction becomes inclusion in the subset intersection, disjunction
inclusion in the subset union, and negation inclusion in the subset
complement negation (Figure \ref{fig:venn_diagrams},
Table \ref{tab:set_truth_tables}).
\begin{figure*}
\centering
\subfigure[]{
\begin{tikzpicture}[scale=0.3, thick]
\draw[color=white] (-0.5, 0) to (20.5, 0);
\draw[color=black] (0, 0) to (20, 0) to (20, 15) to (0, 15) to (0, 0);
\draw[color=gray80, dashed] (12, 9) circle (150pt);
\fill[color=dark] (8, 6) circle (150pt) node[color=white] {$E_{1}$};
\end{tikzpicture}
%
\begin{tikzpicture}[scale=0.3, thick]
\draw[color=white] (-0.5, 0) to (20.5, 0);
\draw[color=black] (0, 0) to (20, 0) to (20, 15) to (0, 15) to (0, 0);
\draw[color=gray80, dashed] (8, 6) circle (150pt);
\fill[color=dark] (12, 9) circle (150pt) node[color=white] {$E_{2}$};
\end{tikzpicture}
}
%
\subfigure[]{
\begin{tikzpicture}[scale=0.3, thick]
\draw[color=black] (0, 0) to (20, 0) to (20, 15) to (0, 15) to (0, 0);
\draw[color=gray80, dashed] (8, 6) circle (150pt);
\draw[color=gray80, dashed] (12, 9) circle (150pt);
\begin{scope}
\clip (8, 6) circle (150pt);
\fill[color=dark] (12, 9) circle (150pt);
\end{scope}
\node[color=white] at (10, 7.5) {$E_{1} \cap E_{2}$};
\end{tikzpicture}
}
%
\subfigure[]{
\begin{tikzpicture}[scale=0.3, thick]
\draw[color=black] (0, 0) to (20, 0) to (20, 15) to (0, 15) to (0, 0);
\fill[color=dark] (8, 6) circle (150pt);
\fill[color=dark] (12, 9) circle (150pt);
\node[color=white] at (10, 7.5) {$E_{1} \cup E_{2}$};
\end{tikzpicture}
}
%
\subfigure[]{
\begin{tikzpicture}[scale=0.3, thick]
\fill[color=dark] (0, 0) to (20, 0) to (20, 15) to (0, 15) to (0, 0);
\draw[color=black] (0, 0) to (20, 0) to (20, 15) to (0, 15) to (0, 0);
\fill[color=white] (8, 6) circle (150pt);
\node[color=white] at (16, 11) {$E_{1}^{c}$};
\end{tikzpicture}
}
%
\caption{In a representation statements are defined as subsets in some
sample space and logical propositions are implemented with set inclusion
or set exclusion. Boolean operations are then implemented with inclusion
in a set complement (negation), set intersection (conjunction), and set
union (disjunction). (a) If two propositions are represented by
inclusion in the sets $E_{1}$ and $E_{2}$, then (b) their conjunction is
represented as inclusion in the set intersection, $E_{1} \cup E_{2}$, and (c)
their disjunction is represented as inclusion in the set union,
$E_{1} \cap E_{2}$. (d) Similarly, the negation of $E_{1}$ is represented
as inclusion in the set complement, $E_{1}^{c}$.}
\label{fig:venn_diagrams}
\end{figure*}
\begin{table}
\centering
\renewcommand{\arraystretch}{1.5}
\begin{tabular}{cc}
\rowcolor[gray]{0.9}
\multicolumn{2}{c}{\textbf{Negation}} \\
%
\rowcolor[gray]{0.9}
$\theta \in E$ & $\theta \in E^{c}$ \\
%
False & True \\
True & False \\
\end{tabular}
%
\vspace{3mm} \\
%
\begin{tabular}{ccc}
\rowcolor[gray]{0.9}
\multicolumn{3}{c}{\textbf{Conjunction}} \\
%
\rowcolor[gray]{0.9}
$\theta \in E_{1}$ & $\theta \in E_{1}$ & $\theta \in E_{1} \cap E_{2}$ \\
%
False & False & False \\
True & False & False \\
False & True & False \\
True & True & True \\
\end{tabular}
%
\hspace{5mm}
%
\begin{tabular}{ccc}
\rowcolor[gray]{0.9}
\multicolumn{3}{c}{\textbf{Disjunction}} \\
%
\rowcolor[gray]{0.9}
$\theta \in E_{1}$ & $\theta \in E_{1}$ & $\theta \in E_{1} \cup E_{2}$ \\
%
False & False & False \\
True & False & True \\
False & True & True \\
True & True & True \\
\end{tabular}
\caption{In a representation statements are defined as subsets in some
sample space and logical propositions are implemented with set inclusion
or set exclusion. Boolean operations are then implemented with inclusion
in a set complement (negation), set intersection (conjunction), and set
union (disjunction), which yield the same truth tables and hence define
a valid representation.}
\label{tab:set_truth_tables}
\end{table}
The same inconsistencies that limited us to consider only well-behaved
logical propositions, however, also limit our representations to only
well-behaved subsets of the sample space. We denote a collection of
subsets of the sample space that is closed under set compliments,
set intersections, and set unions as an \emph{event space}, $\EV{\Theta}$,
with the well-behaved subsets in that space denoted \emph{events}.
Event spaces always include the null event, $E = \emptyset$, which
can never be true as it is always empty. Similarly, event spaces also
always include the trivial event, $E = \Theta$, which can never be false
as every element of the sample space is included in this event.
The pairing of a sample space and an event space is denoted a
\emph{measurable space}, for reasons that will become more clear
in the next chapter.
More formally, then, we should define a representation as a map from
a logical model into a measurable space. In practice this is defined
as a map from the abstract system into the sample space,
%
\begin{equation*}
\mathfrak{r} : \mathcal{S} \rightarrow \Theta,
\end{equation*}
%
that then induces a map from the logical algebra into the event space,
%
\begin{equation*}
\mathfrak{r}_{*} :
\mathcal{L} \! \left( \mathcal{S} \right) \rightarrow \EV{\Theta}.
\end{equation*}
%
A representation is called \emph{faithful} if every logical statement maps
into an event and vice-versa, and Stone's Theorem guarantees that a
faithful representation for any logical model always exists.
For example, descriptions of the planets in our solar system can be
represented with the integers, $\Theta = \left\{1, 2, \ldots, 8 \right\}$,
with each logical proposition represented by inclusion in a subset of
those integers (Figure \ref{fig:discrete_set_logic}). Similarly, the
distance between two objects can be represented by real numbers,
(Figure \ref{fig:real_set_logic}).
\begin{figure*}
\centering
\subfigure[]{
\begin{tikzpicture}[scale=0.3, thick]
\draw[color=white] (-25, 0) to (10, 0);
\node[] at (-15, 0) {The planet has a magnetic field};
\node[] at (7, 2) {$\theta \in E_{1}$};
\fill[color=gray80] (0, 0) circle (25pt) node[color=white] {$\mercury$};
\fill[color=dark] (2, 0) circle (25pt) node[color=white] {$\venus$};
\fill[color=dark] (4, 0) circle (25pt) node[color=white] {$\earth$};
\fill[color=gray80] (6, 0) circle (25pt) node[color=white] {$\mars$};
\fill[color=dark] (8, 0) circle (25pt) node[color=white] {$\jupiter$};
\fill[color=dark] (10, 0) circle (25pt) node[color=white] {$\saturn$};
\fill[color=dark] (12, 0) circle (25pt) node[color=white] {$\uranus$};
\fill[color=dark] (14, 0) circle (25pt) node[color=white] {$\neptune$};
\node[] at (-15, -5) {The planet has moons};
\node[] at (7, -3) {$\theta \in E_{2}$};
\fill[color=gray80] (0, -5) circle (25pt) node[color=white] {$\mercury$};
\fill[color=gray80] (2, -5) circle (25pt) node[color=white] {$\venus$};
\fill[color=dark] (4, -5) circle (25pt) node[color=white] {$\earth$};
\fill[color=dark] (6, -5) circle (25pt) node[color=white] {$\mars$};
\fill[color=dark] (8, -5) circle (25pt) node[color=white] {$\jupiter$};
\fill[color=dark] (10, -5) circle (25pt) node[color=white] {$\saturn$};
\fill[color=dark] (12, -5) circle (25pt) node[color=white] {$\uranus$};
\fill[color=dark] (14, -5) circle (25pt) node[color=white] {$\neptune$};
\end{tikzpicture}
}
%
\subfigure[]{
\begin{tikzpicture}[scale=0.3, thick]
\draw[color=white] (-25, 0) to (10, 0);
\node[align=center] at (-15, 0) {The planet has a magnetic field\\and moons};
\node[] at (7, 2) {$\theta \in E_{1} \cap E_{2}$};
\fill[color=gray80] (0, 0) circle (25pt) node[color=white] {$\mercury$};
\fill[color=gray80] (2, 0) circle (25pt) node[color=white] {$\venus$};
\fill[color=dark] (4, 0) circle (25pt) node[color=white] {$\earth$};
\fill[color=gray80] (6, 0) circle (25pt) node[color=white] {$\mars$};
\fill[color=dark] (8, 0) circle (25pt) node[color=white] {$\jupiter$};
\fill[color=dark] (10, 0) circle (25pt) node[color=white] {$\saturn$};
\fill[color=dark] (12, 0) circle (25pt) node[color=white] {$\uranus$};
\fill[color=dark] (14, 0) circle (25pt) node[color=white] {$\neptune$};
\end{tikzpicture}
}
%
\subfigure[]{
\begin{tikzpicture}[scale=0.3, thick]
\draw[color=white] (-25, 0) to (10, 0);
\node[align=center] at (-15, 0) {The planet has a magnetic field\\or moons};
\node[] at (7, 2) {$\theta \in E_{1} \cup E_{2}$};
\fill[color=gray80] (0, 0) circle (25pt) node[color=white] {$\mercury$};
\fill[color=dark] (2, 0) circle (25pt) node[color=white] {$\venus$};
\fill[color=dark] (4, 0) circle (25pt) node[color=white] {$\earth$};
\fill[color=dark] (6, 0) circle (25pt) node[color=white] {$\mars$};
\fill[color=dark] (8, 0) circle (25pt) node[color=white] {$\jupiter$};
\fill[color=dark] (10, 0) circle (25pt) node[color=white] {$\saturn$};
\fill[color=dark] (12, 0) circle (25pt) node[color=white] {$\uranus$};
\fill[color=dark] (14, 0) circle (25pt) node[color=white] {$\neptune$};
\end{tikzpicture}
}
%
\subfigure[]{
\begin{tikzpicture}[scale=0.3, thick]
\draw[color=white] (-25, 0) to (10, 0);
\node[align=center] at (-15, 0) {The planet does not\\have a magnetic field};
\node[] at (7, 2) {$\theta \in E_{1}^{c}$};
\fill[color=dark] (0, 0) circle (25pt) node[color=white] {$\mercury$};
\fill[color=gray80] (2, 0) circle (25pt) node[color=white] {$\venus$};
\fill[color=gray80] (4, 0) circle (25pt) node[color=white] {$\earth$};
\fill[color=dark] (6, 0) circle (25pt) node[color=white] {$\mars$};
\fill[color=gray80] (8, 0) circle (25pt) node[color=white] {$\jupiter$};
\fill[color=gray80] (10, 0) circle (25pt) node[color=white] {$\saturn$};
\fill[color=gray80] (12, 0) circle (25pt) node[color=white] {$\uranus$};
\fill[color=gray80] (14, 0) circle (25pt) node[color=white] {$\neptune$};
\end{tikzpicture}
}
\caption{(a) Systems that take discrete values, such as the planets in
our solar system, can be represented with integers, with any logical
proposition about the planets represented by inclusion in a subset of
integers. (b) Conjuction of logical propositions is implemented with
inclusion in set intersections, (c) disjunction with inclusion in set unions,
and (d) negation with inclusion in set complements.}
\label{fig:discrete_set_logic}
\end{figure*}
\begin{figure*}
\centering
\subfigure[]{
\begin{tikzpicture}[scale=0.3, thick]
\draw[color=white] (-27, 0) to (17, 0);
\node[align=center] at (-15, 0) {The distance is\\less than five inches};
\node[] at (7.5, 2) {$\theta \in E_{1}$};
\draw[|->] (0, 0) -- (14,0) node[right] {$x$};
\draw[line width=1mm, color=dark] (0, 0) node[] {$\,($} -- (5, 0) node[] {$\!)$};
\node[align=center] at (-15, -5) {The distance is between\\three and seven inches};
\node[] at (7.5, -3) {$\theta \in E_{2}$};
\draw[|->] (0, -5) -- (14,-5) node[right] {$x$};
\draw[line width=1mm, color=dark] (3, -5) node[] {$\,($} -- (7,-5) node[] {$\!)$};
\end{tikzpicture}
}
%
\subfigure[]{
\begin{tikzpicture}[scale=0.3, thick]
\draw[color=white] (-27, 0) to (17, 0);
\node[align=center] at (-15, 0) {The distance is less than five inches\\and
between three and seven inches};
\node[] at (7.5, 2) {$\theta \in E_{1} \cap E_{2}$};
\draw[|->] (0, 0) -- (14,0) node[right] {$x$};
\draw[line width=1mm, color=dark] (3, 0) node[] {$\,($} -- (5, 0) node[] {$\!)$};
\end{tikzpicture}
}
%
\subfigure[]{
\begin{tikzpicture}[scale=0.3, thick]
\draw[color=white] (-27, 0) to (17, 0);
\node[align=center] at (-15, 0) {The distance is less than five inches\\or
between three and seven inches};
\node[] at (7.5, 2) {$\theta \in E_{1} \cup E_{2}$};
\draw[|->] (0, 0) -- (14, 0) node[right] {$x$};
\draw[line width=1mm, color=dark] (0, 0) node[] {$\,($} -- (7, 0) node[] {$\!)$};
\end{tikzpicture}
}
%
\subfigure[]{
\begin{tikzpicture}[scale=0.3, thick]
\draw[color=white] (-27, 0) to (17, 0);
\node[align=center] at (-15, 0) {The distance is not\\less than five inches};
\node[] at (7.5, 2) {$\theta \in E_{1}^{c}$};
\draw[|->] (0, 0) -- (14, 0) node[right] {$x$};
\draw[line width=1mm, color=dark] (5, 0) node[] {$\,($} -- (13, 0);
\end{tikzpicture}
}
\caption{(a) Systems that take continuous values, such as distances,
can be represented with real numbers, with any logical proposition
represented by inclusion in well-behaved subsets of real numbers.
(b) Conjuction of logical propositions is implemented with inclusion
in set intersections, (c) disjunction with inclusion in set unions, and
(d) negation with inclusion in set complements.}
\label{fig:real_set_logic}
\end{figure*}
\section{Equivalent Representations}
Because they map abstract systems into explicit contexts, representations
are critical to the practical implementation of logical models and their
manipulations. Unfortunately representations also bring with them a
dangerous subtlety -- there is no \emph{unique} representation of a
given abstract system.
A map from one sample space into another, $s : \Theta \rightarrow \Omega$,
is \emph{measurable} if every event in $\Omega$ is generated by an event
in $\Theta$,
%
\begin{equation*}
s^{-1} \! \left( E_{\Omega} \right) \in \EV{\Theta}.
\end{equation*}
%
Measurable maps allow us to completely generate the measurable space
$\left( \Omega, \EV{\Omega} \right)$ by mapping forward elements of the
initial measurable space $\left(\Theta, \EV{\Theta} \right)$; we cannot,
however, necessarily achieve the converse as there is no guarantee that
every event in $\EV{\Theta}$ maps to an event in $\EV{\Omega}$.
This process is also known as \emph{pushing forward} the measurable
space $\left( \Theta, \EV{\Theta} \right)$ into the measurable space
$\left( \Omega, \EV{\Omega} \right)$, and in this context
$\left( \Omega, \EV{\Omega} \right)$ is denoted the \emph{pushforward}
of $\left( \Theta, \EV{\Theta} \right)$ by $s$.
Composing a faithful representation with a measurable map,
%
\begin{equation*}
s \circ \mathfrak{e} : \mathcal{S} \rightarrow \Theta \rightarrow \Omega,
\end{equation*}
%
yields another valid representation but not necessarily a faithful one.
Because there is no guarantee that every event in $\EV{\Theta}$ maps
to an event in $\EV{\Omega}$, this pushforward may be able to represent
only a handful of propositions in the logical model. For example,
a measurable map might round real numbers down to discrete numbers,
losing the ability to represent all but a countable number of logical
propositions.
In order to maintain a faithful representation we need to ensure that
the events spaces are preserved by both the map and its inverse. Here
we will say that a map $s : \Theta \rightarrow \Omega$ is
\emph{doubly-measurable} if every event in $\EV{\Theta}$ maps to an
event in $\EV{\Omega}$ and vice versa,
%
\begin{align*}
s \! \left( E_{\Theta} \right) &\in \EV{\Omega}
\\
s^{-1} \! \left( E_{\Omega} \right) &\in \EV{\Theta}.
\end{align*}
%
A doubly-measurable map allows us to recover either measurable space
as a pushforward of the other, and the composition of a faithful
representation with a doubly-measurable map,
%
\begin{equation*}
s \circ \mathfrak{e} : \mathcal{S} \rightarrow \Theta \rightarrow \Omega,
\end{equation*}
%
is then another faithful representation. We warn the reader that there
is no conventional terminology for these maps and we have introduced
doubly-measurable here as our own terminology.
When a doubly-measurable map exists between two measurable spaces
then the two spaces can be used to specify the same logical model,
and we say that they are \emph{equivalent representations} as they provide
equivalent descriptions of a given system
(Table \ref{tab:representation_examples}). The set of measurable spaces
equivalent to $\left( \Theta, \EV{\Theta} \right)$ will be denoted
$\lfloor \Theta, \EV{\Theta} \rfloor$.
For example, if the sample space is discrete then we can always define
an equivalent measurable space by simply permuting the labels. It doesn't
matter if we order the planets by distance from the sun, by diameter, or
by any other metric: the measurable spaces quantify the same information
(Figure \ref{fig:permuting_discrete_spaces}). Similarly, we can always
apply a transformation that warps real numbers to map between real
sample spaces without compromising our ability to represent propositions
with inclusion in events. One of the most common ways this manifests in
practice is when our descriptions require units. The information we quantify
doesn't depend on whether we express distance in {\aa}ngstr\"{o}ms or
inches or meters or furlongs: each unit defines a separate but equivalent
sample space.
\begin{figure*}
\centering
\begin{tikzpicture}[scale=0.3, thick]
\foreach \i in {1, 2, ..., 8} {
\node[] at ({2 * \i - 2}, 2) {$\i$};
}
\fill[color=dark] (0, 0) circle (25pt) node[color=white] {$\mercury$};
\fill[color=dark] (2, 0) circle (25pt) node[color=white] {$\venus$};
\fill[color=dark] (4, 0) circle (25pt) node[color=white] {$\earth$};
\fill[color=dark] (6, 0) circle (25pt) node[color=white] {$\mars$};
\fill[color=dark] (8, 0) circle (25pt) node[color=white] {$\jupiter$};
\fill[color=dark] (10, 0) circle (25pt) node[color=white] {$\saturn$};
\fill[color=dark] (12, 0) circle (25pt) node[color=white] {$\uranus$};
\fill[color=dark] (14, 0) circle (25pt) node[color=white] {$\neptune$};
\draw[->] (0, -1) -- (0, -5);
\draw[->] (2, -1) -- (4, -5);
\draw[->] (4, -1) -- (6, -5);
\draw[->] (6, -1) -- (2, -5);
\draw[->] (8, -1) -- (14, -5);
\draw[->] (10, -1) -- (12, -5);
\draw[->] (12, -1) -- (10, -5);
\draw[->] (14, -1) -- (8, -5);
\fill[color=dark] (0, -6) circle (25pt) node[color=white] {$\mercury$};
\fill[color=dark] (2, -6) circle (25pt) node[color=white] {$\mars$};
\fill[color=dark] (4, -6) circle (25pt) node[color=white] {$\venus$};
\fill[color=dark] (6, -6) circle (25pt) node[color=white] {$\earth$};
\fill[color=dark] (8, -6) circle (25pt) node[color=white] {$\neptune$};
\fill[color=dark] (10, -6) circle (25pt) node[color=white] {$\uranus$};
\fill[color=dark] (12, -6) circle (25pt) node[color=white] {$\saturn$};
\fill[color=dark] (14, -6) circle (25pt) node[color=white] {$\jupiter$};
\foreach \i in {1, 2, ..., 8} {
\node[] at ({2 * \i - 2}, -8) {$\i$};
}
\end{tikzpicture}
\caption{Discrete sample spaces are not unique way to express a system
as we can always permute the arbitrary labels to yield an equivalent sample
space. For example, we could order the planets by distance from the sun
or diameter without affecting our ability to describe the planets themselves.}
\label{fig:permuting_discrete_spaces}
\end{figure*}
Ultimately, there are many ways to describe of a given logical model and
measurable maps translate from one equivalent representation to another.
Although certain representations may be more useful than others in practice,
we have to be careful to ensure that our final analysis does not depend on
the irrelevant details of any particular representation.
\section{Implications}
There are various Boolean operations that we can derive from the basic
operations of negation, conjunction, and disjunction, and one of the most
powerful are \emph{implications}, which assert the truth of a statement
conditioned on the truth of another statement. For example, we might not
be able to assert that ``a planet has a magnetic field'' universally, but we
might be able to assert the implication ``if a planet has a rotating core then
it has a magnetic field''.
Implications are particularly naturally when we want to describe a target
system conditioned on the state of some other system. Letting
$\left(\Theta, \EV{\Theta} \right)$ be a measurable space capable of
representing the target system and $\left(\Phi, \EV{\Phi} \right)$ be a
measurable space capable of representing the conditioning system, an
implication is represented as a map from points in $\Phi$ to events in
$\EV{\Theta}$,
%
\begin{align*}
\iota_{\Theta \mid \Phi} :& \, \Phi \rightarrow \EV{\Theta}
\\
& \, \phi \mapsto \iota_{\Theta \mid \Phi} \! \left( \phi \right).
\end{align*}
%
For any event in the conditioning space, $E_{\Phi} \in \EV{\Phi}$,
the implication also defines an event in the joint sample space,
$\Theta \times \Phi$, as the union of all of the events implied by each
point in $E_{\Phi}$,
%
\begin{align*}
\iota^{*}_{\Theta \mid \Phi}
&: \EV{\Phi} \rightarrow \EV{\Theta \times \Phi}
\\
& \quad \; E_{\Phi} \; \mapsto
\bigcup_{\phi \in E_{\Phi} } \iota_{\Theta \mid \Phi} \! \left( \phi \right).
\end{align*}
Because relationships are often much easier to reason about than the
entirety of a complex system, these induced events can drastically facilitate
the description of complex systems by decomposing logical propositions
into simpler implications. For example, instead of having to enumerate all
possible combinations of planetary core behavior and magnetic field, we
can utilize the single implication that only a rotating core generates a
magnetic field to build any logical proposition about the joint core-magnetic
field system from a logical proposition about the core alone.
More formally, any logical proposition representation in a joint sample space
$\prod_{n = 1}^{N} \Theta_{n}$ can be built up sequentially, starting with an
event in one component and adding implications one after another,
%
\begin{align*}
& E_{\Theta_{1}} \\
& \iota^{*}_{\Theta_{2} \mid \Theta_{1}} \! \left( E_{\Theta_{1}} \right) \\
& \iota^{*}_{\Theta_{3} \mid \Theta_{2}, \Theta_{1}}
\circ
\iota^{*}_{\Theta_{2} \mid \Theta_{1}} \! \left( E_{\Theta_{1}} \right) \\
& \cdots \\
& \iota^{*}_{\Theta_{N} \mid \Theta_{N - 1}, \ldots, \Theta_{2}, \Theta_{1}}
\circ
\iota^{*}_{\Theta_{3} \mid \Theta_{2}, \Theta_{1}}
\circ
\iota^{*}_{\Theta_{2} \mid \Theta_{1}} \! \left( E_{\Theta_{1}} \right).
\end{align*}