formalizing exponents 3 and 4 of fermat’s last …€™s last theorem using the proof-assistant...
TRANSCRIPT
![Page 1: Formalizing exponents 3 and 4 of Fermat’s Last …€™s Last Theorem using the proof-assistant ‘Isabelle’ Roelof Oosterhuis Supervisors: Jaap Top & Wim Hesselink DIAMANT …](https://reader034.vdocuments.us/reader034/viewer/2022051800/5ad10ae07f8b9a4e7a8ea287/html5/thumbnails/1.jpg)
1This information is confidential and was prepared by Bain & Company solely for the use of our client; it is not to be relied on by any 3rd party without Bain's prior written consent.
AMS
Slide design copied from
Formalizing exponents 3 and 4 of
Fermat’s Last Theorem using
the proof-assistant ‘Isabelle’
Roelof OosterhuisSupervisors: Jaap Top & Wim Hesselink
DIAMANT symposium 30 November 2007 (14.25-14.45)
![Page 2: Formalizing exponents 3 and 4 of Fermat’s Last …€™s Last Theorem using the proof-assistant ‘Isabelle’ Roelof Oosterhuis Supervisors: Jaap Top & Wim Hesselink DIAMANT …](https://reader034.vdocuments.us/reader034/viewer/2022051800/5ad10ae07f8b9a4e7a8ea287/html5/thumbnails/2.jpg)
2This information is confidential and was prepared by Bain & Company solely for the use of our client; it is not to be relied on by any 3rd party without Bain's prior written consent.
AMS
Agenda
• Aim of my research
• Proof assistants à Isabelle
• Fermat’s last theorem- case n=4- case n=3
• Results of my research- space factor- time factor
• Discussion
![Page 3: Formalizing exponents 3 and 4 of Fermat’s Last …€™s Last Theorem using the proof-assistant ‘Isabelle’ Roelof Oosterhuis Supervisors: Jaap Top & Wim Hesselink DIAMANT …](https://reader034.vdocuments.us/reader034/viewer/2022051800/5ad10ae07f8b9a4e7a8ea287/html5/thumbnails/3.jpg)
3This information is confidential and was prepared by Bain & Company solely for the use of our client; it is not to be relied on by any 3rd party without Bain's prior written consent.
AMS
Pilot study for a larger ideal
Aims of my research:
• To give a formal proof of FLT3
• How getting started with ‘formalizing mathematics’?- How to choose a proof assistant?- How getting familiar with such a program?
• How does a formalization work in practice?- What problems does one encounter?- How much time does it take?- How ‘doable’ are problems like FLT3&4?
“Formalize and verify by computer a proof of Fermat’s Last Theorem.”
Prof. dr. Jan Bergstra(nr 1. of his list of ‘ten challenging research problems for computer science’)
![Page 4: Formalizing exponents 3 and 4 of Fermat’s Last …€™s Last Theorem using the proof-assistant ‘Isabelle’ Roelof Oosterhuis Supervisors: Jaap Top & Wim Hesselink DIAMANT …](https://reader034.vdocuments.us/reader034/viewer/2022051800/5ad10ae07f8b9a4e7a8ea287/html5/thumbnails/4.jpg)
4This information is confidential and was prepared by Bain & Company solely for the use of our client; it is not to be relied on by any 3rd party without Bain's prior written consent.
AMS
Agenda
• Aim of my research
• Proof assistants à Isabelle
• Fermat’s last theorem- case n=4- case n=3
• Results of my research- space factor- time factor
• Discussion
![Page 5: Formalizing exponents 3 and 4 of Fermat’s Last …€™s Last Theorem using the proof-assistant ‘Isabelle’ Roelof Oosterhuis Supervisors: Jaap Top & Wim Hesselink DIAMANT …](https://reader034.vdocuments.us/reader034/viewer/2022051800/5ad10ae07f8b9a4e7a8ea287/html5/thumbnails/5.jpg)
5This information is confidential and was prepared by Bain & Company solely for the use of our client; it is not to be relied on by any 3rd party without Bain's prior written consent.
AMS
Formalizing mathematics: what? why?
• Formalizing mathematics = expressing statements and proofs in a usually small and simple formal language with strict rules of grammar and unambiguous semantics
• Opportunities when combined with computer science:-Proof-checking can be automated and can be more reliable
(only the checking program has to be checked)-Proofs are completely explicit and highly accessible-Opportunities for better online collaboration
(compare with wikipedia)-Computer can do laborious parts
(many case distinctions, calculations, ..)
![Page 6: Formalizing exponents 3 and 4 of Fermat’s Last …€™s Last Theorem using the proof-assistant ‘Isabelle’ Roelof Oosterhuis Supervisors: Jaap Top & Wim Hesselink DIAMANT …](https://reader034.vdocuments.us/reader034/viewer/2022051800/5ad10ae07f8b9a4e7a8ea287/html5/thumbnails/6.jpg)
6This information is confidential and was prepared by Bain & Company solely for the use of our client; it is not to be relied on by any 3rd party without Bain's prior written consent.
AMS
Results in mechanised proving
• Four Colour Theorem (2004). By Georges Gonthier using Coq (60,000 lines).
• Prime Number Theorem (2005).By Jeremy Avigad + students using Isabelle (30,000 lines)
• Flyspeck Project (not finished).By several research groups / proof assistants.
• Large part of undergraduate mathematics
![Page 7: Formalizing exponents 3 and 4 of Fermat’s Last …€™s Last Theorem using the proof-assistant ‘Isabelle’ Roelof Oosterhuis Supervisors: Jaap Top & Wim Hesselink DIAMANT …](https://reader034.vdocuments.us/reader034/viewer/2022051800/5ad10ae07f8b9a4e7a8ea287/html5/thumbnails/7.jpg)
7This information is confidential and was prepared by Bain & Company solely for the use of our client; it is not to be relied on by any 3rd party without Bain's prior written consent.
AMS
Proof assistants: Coq
Source: http://www.cs.ru.nl/~freek/100
![Page 8: Formalizing exponents 3 and 4 of Fermat’s Last …€™s Last Theorem using the proof-assistant ‘Isabelle’ Roelof Oosterhuis Supervisors: Jaap Top & Wim Hesselink DIAMANT …](https://reader034.vdocuments.us/reader034/viewer/2022051800/5ad10ae07f8b9a4e7a8ea287/html5/thumbnails/8.jpg)
8This information is confidential and was prepared by Bain & Company solely for the use of our client; it is not to be relied on by any 3rd party without Bain's prior written consent.
AMS
Proof assistants: Mizar
![Page 9: Formalizing exponents 3 and 4 of Fermat’s Last …€™s Last Theorem using the proof-assistant ‘Isabelle’ Roelof Oosterhuis Supervisors: Jaap Top & Wim Hesselink DIAMANT …](https://reader034.vdocuments.us/reader034/viewer/2022051800/5ad10ae07f8b9a4e7a8ea287/html5/thumbnails/9.jpg)
9This information is confidential and was prepared by Bain & Company solely for the use of our client; it is not to be relied on by any 3rd party without Bain's prior written consent.
AMS
Proof assistants: HOL-Light
![Page 10: Formalizing exponents 3 and 4 of Fermat’s Last …€™s Last Theorem using the proof-assistant ‘Isabelle’ Roelof Oosterhuis Supervisors: Jaap Top & Wim Hesselink DIAMANT …](https://reader034.vdocuments.us/reader034/viewer/2022051800/5ad10ae07f8b9a4e7a8ea287/html5/thumbnails/10.jpg)
10This information is confidential and was prepared by Bain & Company solely for the use of our client; it is not to be relied on by any 3rd party without Bain's prior written consent.
AMS
Proof assistants: Isabelle
![Page 11: Formalizing exponents 3 and 4 of Fermat’s Last …€™s Last Theorem using the proof-assistant ‘Isabelle’ Roelof Oosterhuis Supervisors: Jaap Top & Wim Hesselink DIAMANT …](https://reader034.vdocuments.us/reader034/viewer/2022051800/5ad10ae07f8b9a4e7a8ea287/html5/thumbnails/11.jpg)
11This information is confidential and was prepared by Bain & Company solely for the use of our client; it is not to be relied on by any 3rd party without Bain's prior written consent.
AMS
Isabelle characteristics
• Readable i/o
• Good documentation
• Several logics
• Two input formats (Isabelle & Isar)
• Isabelle’s automation is good at:-Logical reasoning-Calculating with equalities, like
(at least if you know the commands)-Not: calculating with inequalities (perhaps no prover is), like
![Page 12: Formalizing exponents 3 and 4 of Fermat’s Last …€™s Last Theorem using the proof-assistant ‘Isabelle’ Roelof Oosterhuis Supervisors: Jaap Top & Wim Hesselink DIAMANT …](https://reader034.vdocuments.us/reader034/viewer/2022051800/5ad10ae07f8b9a4e7a8ea287/html5/thumbnails/12.jpg)
12This information is confidential and was prepared by Bain & Company solely for the use of our client; it is not to be relied on by any 3rd party without Bain's prior written consent.
AMS
Agenda
• Aim of my research
• Proof assistants à Isabelle
• Fermat’s last theorem- case n=4- case n=3
• Results of my research- space factor- time factor
• Discussion
![Page 13: Formalizing exponents 3 and 4 of Fermat’s Last …€™s Last Theorem using the proof-assistant ‘Isabelle’ Roelof Oosterhuis Supervisors: Jaap Top & Wim Hesselink DIAMANT …](https://reader034.vdocuments.us/reader034/viewer/2022051800/5ad10ae07f8b9a4e7a8ea287/html5/thumbnails/13.jpg)
13This information is confidential and was prepared by Bain & Company solely for the use of our client; it is not to be relied on by any 3rd party without Bain's prior written consent.
AMS
Fermat’s Last Theorem
• Fermat proved the case n = 4 (using infinite descent and Euclid’s construction of Pythagorean triples)
• Euler (1707-1783) proved the case n = 3(two versions, both contain errors)
• Wiles proved (in 1995) the case n is a prime ≥ 5(after several smaller and bigger repairs)
This completed the proof of FLT:
![Page 14: Formalizing exponents 3 and 4 of Fermat’s Last …€™s Last Theorem using the proof-assistant ‘Isabelle’ Roelof Oosterhuis Supervisors: Jaap Top & Wim Hesselink DIAMANT …](https://reader034.vdocuments.us/reader034/viewer/2022051800/5ad10ae07f8b9a4e7a8ea287/html5/thumbnails/14.jpg)
14This information is confidential and was prepared by Bain & Company solely for the use of our client; it is not to be relied on by any 3rd party without Bain's prior written consent.
AMS
The easy case n=4: informal proof
![Page 15: Formalizing exponents 3 and 4 of Fermat’s Last …€™s Last Theorem using the proof-assistant ‘Isabelle’ Roelof Oosterhuis Supervisors: Jaap Top & Wim Hesselink DIAMANT …](https://reader034.vdocuments.us/reader034/viewer/2022051800/5ad10ae07f8b9a4e7a8ea287/html5/thumbnails/15.jpg)
15This information is confidential and was prepared by Bain & Company solely for the use of our client; it is not to be relied on by any 3rd party without Bain's prior written consent.
AMS
Example of Isabelleinteraction
![Page 16: Formalizing exponents 3 and 4 of Fermat’s Last …€™s Last Theorem using the proof-assistant ‘Isabelle’ Roelof Oosterhuis Supervisors: Jaap Top & Wim Hesselink DIAMANT …](https://reader034.vdocuments.us/reader034/viewer/2022051800/5ad10ae07f8b9a4e7a8ea287/html5/thumbnails/16.jpg)
16This information is confidential and was prepared by Bain & Company solely for the use of our client; it is not to be relied on by any 3rd party without Bain's prior written consent.
AMS
The tricky case n=3…
Source: http://www.fermatslasttheorem.blogspot.com (25-11-’07)
![Page 17: Formalizing exponents 3 and 4 of Fermat’s Last …€™s Last Theorem using the proof-assistant ‘Isabelle’ Roelof Oosterhuis Supervisors: Jaap Top & Wim Hesselink DIAMANT …](https://reader034.vdocuments.us/reader034/viewer/2022051800/5ad10ae07f8b9a4e7a8ea287/html5/thumbnails/17.jpg)
17This information is confidential and was prepared by Bain & Company solely for the use of our client; it is not to be relied on by any 3rd party without Bain's prior written consent.
AMS
Agenda
• Aim of my research
• Proof assistants à Isabelle
• Fermat’s last theorem- case n=4- case n=3
• Results of my research- space factor- time factor
• Discussion
![Page 18: Formalizing exponents 3 and 4 of Fermat’s Last …€™s Last Theorem using the proof-assistant ‘Isabelle’ Roelof Oosterhuis Supervisors: Jaap Top & Wim Hesselink DIAMANT …](https://reader034.vdocuments.us/reader034/viewer/2022051800/5ad10ae07f8b9a4e7a8ea287/html5/thumbnails/18.jpg)
18This information is confidential and was prepared by Bain & Company solely for the use of our client; it is not to be relied on by any 3rd party without Bain's prior written consent.
AMS
Results (1)
• Formal proof of FLT4: 1000 lines
• Formal proof of FLT3: 2500 (extra) lines
• Formal proof of Lagrange’s Four-square Theorem (“any natural number can be written as the sum of 4 squares”): 500 lines, 10 hours
How much ‘more’ work involves this, compared with
‘informal’ mathematics?
![Page 19: Formalizing exponents 3 and 4 of Fermat’s Last …€™s Last Theorem using the proof-assistant ‘Isabelle’ Roelof Oosterhuis Supervisors: Jaap Top & Wim Hesselink DIAMANT …](https://reader034.vdocuments.us/reader034/viewer/2022051800/5ad10ae07f8b9a4e7a8ea287/html5/thumbnails/19.jpg)
19This information is confidential and was prepared by Bain & Company solely for the use of our client; it is not to be relied on by any 3rd party without Bain's prior written consent.
AMS
Space factor: 4 - 20
0
2
4
6
8
10
Pages required for a proof ofLagrange’s Four-square theorem
Isabelle
main proof
small lemma's
10.0
First-year's handout
2.5
Master's thesis
0.5
![Page 20: Formalizing exponents 3 and 4 of Fermat’s Last …€™s Last Theorem using the proof-assistant ‘Isabelle’ Roelof Oosterhuis Supervisors: Jaap Top & Wim Hesselink DIAMANT …](https://reader034.vdocuments.us/reader034/viewer/2022051800/5ad10ae07f8b9a4e7a8ea287/html5/thumbnails/20.jpg)
20This information is confidential and was prepared by Bain & Company solely for the use of our client; it is not to be relied on by any 3rd party without Bain's prior written consent.
AMS
Time factor: 2 - 5
0
2
4
6
8
10
Hours required for producing a proof ofLagrange’s Four-square theorem
Isabelle
make detailedversion
study proof
inspect 'priorknowledge'
improvements
raw version ofproof
10
First-year's handout
5
Master's thesis
2
![Page 21: Formalizing exponents 3 and 4 of Fermat’s Last …€™s Last Theorem using the proof-assistant ‘Isabelle’ Roelof Oosterhuis Supervisors: Jaap Top & Wim Hesselink DIAMANT …](https://reader034.vdocuments.us/reader034/viewer/2022051800/5ad10ae07f8b9a4e7a8ea287/html5/thumbnails/21.jpg)
21This information is confidential and was prepared by Bain & Company solely for the use of our client; it is not to be relied on by any 3rd party without Bain's prior written consent.
AMS
Agenda
• Aim of my research
• Proof assistants à Isabelle
• Fermat’s last theorem- case n=4- case n=3
• Results of my research- space factor- time factor
• Discussion
![Page 22: Formalizing exponents 3 and 4 of Fermat’s Last …€™s Last Theorem using the proof-assistant ‘Isabelle’ Roelof Oosterhuis Supervisors: Jaap Top & Wim Hesselink DIAMANT …](https://reader034.vdocuments.us/reader034/viewer/2022051800/5ad10ae07f8b9a4e7a8ea287/html5/thumbnails/22.jpg)
22This information is confidential and was prepared by Bain & Company solely for the use of our client; it is not to be relied on by any 3rd party without Bain's prior written consent.
AMS
The battle is not over yet…
32
33
34
35
36
37
38
39
40
41
Number of formalized theoremsfrom 'top 100'
1-3-07 15-7-07 26-11-07
Isabelle
Mizar
ProofPower
Coq
HOL-Light
NB: HOL-Light omitted in graphSource: Freek Wiedijk, http://www.cs.ru.nl/~freek/100
![Page 23: Formalizing exponents 3 and 4 of Fermat’s Last …€™s Last Theorem using the proof-assistant ‘Isabelle’ Roelof Oosterhuis Supervisors: Jaap Top & Wim Hesselink DIAMANT …](https://reader034.vdocuments.us/reader034/viewer/2022051800/5ad10ae07f8b9a4e7a8ea287/html5/thumbnails/23.jpg)
23This information is confidential and was prepared by Bain & Company solely for the use of our client; it is not to be relied on by any 3rd party without Bain's prior written consent.
AMS
Collaborate now to speed up development
Progression in development of theorem provers(a little suggestive)
0 1 2 3 4 5 6 7 8
Developone prover
'Survival ofthe fittest'
![Page 24: Formalizing exponents 3 and 4 of Fermat’s Last …€™s Last Theorem using the proof-assistant ‘Isabelle’ Roelof Oosterhuis Supervisors: Jaap Top & Wim Hesselink DIAMANT …](https://reader034.vdocuments.us/reader034/viewer/2022051800/5ad10ae07f8b9a4e7a8ea287/html5/thumbnails/24.jpg)
24This information is confidential and was prepared by Bain & Company solely for the use of our client; it is not to be relied on by any 3rd party without Bain's prior written consent.
AMS
Collaborate now to prevent doing work twice
• None of the current systems (Mizar, Isabelle, HOL, ProofPower, Coq) is acceptable as the QED system yet
(F.Wiedijk, ‘The QED manifesto revisited’, http://mizar.org/trybulec65/8.pdf)
• You don’t want to formalize Wiles’ proof twice…
![Page 25: Formalizing exponents 3 and 4 of Fermat’s Last …€™s Last Theorem using the proof-assistant ‘Isabelle’ Roelof Oosterhuis Supervisors: Jaap Top & Wim Hesselink DIAMANT …](https://reader034.vdocuments.us/reader034/viewer/2022051800/5ad10ae07f8b9a4e7a8ea287/html5/thumbnails/25.jpg)
25This information is confidential and was prepared by Bain & Company solely for the use of our client; it is not to be relied on by any 3rd party without Bain's prior written consent.
AMS
Discussion
• Mechanized theorem proving / proof verification…
-is not that far away from the usual mathematical work (anymore)
-will gain an important role in the daily life of mathematical research and education, within a few decades
-is an accessible research field, for mathematicians as well as for computer scientists
-requires immediate world wide collaboration8 to speed up the development8 to prevent doing proofs twice