20181210, 22:46  #1 
Feb 2018
1 Posts 
Computing Dedekind number D(9).
10Dec2018
Unpublished work, but privated registered. Hello, world. Im computing Dedekind numbers. Using a desktop i computed D(7). But i find many lines of research. I tell you what i find, here.  I extended the function to n. F(n): 2,(3) ,5, (6), 11,14,19, (20), 39... Sloane A132581.  Algorithms to evaluate F(n) and get the set C(n).  full or sequential chained.  F(for 2 powers, the Dedekind numbers).  using the published methods, someones extended.  many new methods: using rules over sets, trees, graphs or patterns.  evaluate: F(kn) from C(n).  get set C(kn) from C(n).  a compressed form for sets C(n), using symbol "X" as (0 and 1). Cardinality of set is few times less than F(). With good treelike sizes. I developped the related 3valued tables to operate with xcompressed sets. Example. C(4), Dedekind set n 2, is: 0000 0XX1 1111 A card 6, on 3 elements. C(6), 14 on 6. D(5), F(32), 7581 on 1200.  I find properties of F(n) that enables computing shortcuts to D. numbers. Examples. D(4)=F( 16)= 2*F(14)=2*84 D(8)=F(256)= 2*F(240). D(8) evaluation: (x2)C(120), (x3)C(80), (x4)C(60).  Properties of Differences(n) ¿ is a way to explicit expression? SLOANE. A132582.  using 3value logic, i find: F(n,3) = F(2n,2) D(7)=F(128,2)=F(64,3).  F(n fixed,s) give a polynomial n graded on s. I call them Pol. of Dedekind. Table F(n,s). 1. 1,2,3,... 2. 1,3,6,10,.. 3. 1,5,14,30,55,91,140,.. 4. 1,6,20,50,105,.. 6. 1,14,84,330,1001,.. 8. 1,20,168,.. Questions.  whats new, whats old.  whats seams useful, whats not.  Ways to try to find explicit form for F(n).  One alg. to reduce C(n) to minimum cardinality.  ¿someone computing D(9)? Please, comment. 
20181211, 03:39  #2 
Aug 2006
3^{2}·5·7·19 Posts 
TL;DR: The OP is trying to compute D(9) with D(n) = A000372(n) = F(2^n) = A132581(2^n). See the sequences for background and references.
Given the growth of the sequence (each term seems to be about the square of the previous term) I expect the computation to be pretty intense. I'd love to see calculation ideas. 
20181211, 15:22  #3 
Feb 2018
2^{5}·3 Posts 
What means TL , DR, OP ?
" I'd love to see calculation ideas"
Well, Im computing D(n) for last 4 months. Today best options i have are xcompressed sets and computing shorcuts. I find many others interesting ways but i cant explore them. Time. JMMA Note: I get some forum account problems. I post last work version. Title . Comments on some properties of Dedekind numbers. 11Dec2018. JMMA. Work private registered. Im computing Dedekind numbers. Using a desktop i computed D(7). And i find many lines of research.  I extended the function to n. F(n): 2,(3),5,(6),11,14,19,(20) ,39..84..(168)..2008..(7581).. Sloane A132581.  Algorithms to evaluate F(n) and get set C(n).  full or sequentialchained.  F(for 2 powers): the Dedekind numbers.  using the published methods, someones extended.  OR using new methods: rules over sets, trees, graphs or patterns.  rule based from condition (a LE b).  evaluate: F(kn) from C(n).  get set C(kn) from C(n). Example. D(8)=(x2)C(120)= (x3)C(80)= (x4)C(60).  a compressed form for sets C(n), using symbol "X" as (0 and 1). Cardinality of set is few times less than F(). With good treelike sizes. I developped the related 3valued tables to operate with xcompressed sets. Examples. Set C(4), Dedekind set n 2, is: 0000 0XX1 1111 A card 6, on 3 elements. Set C(6) 0000XX 0010X1 01001X 011011 0XX111 111111 Card 14, on 6. D(5)= F(32)= 7581, C(32) on 1200. Table (a LE b) for xcompressed sets.  0 1 X 0 0 1 2 1 0 1 1 X 1 2 3  Similar tables to get C(2n).  I find some properties of F(n). F(2^e)= F(2^e  2^d) + F( ((2^d1)/2^d)*(2^e)) D(4)=F( 16)= 2*F(14)=2*84 D(8)=F(256)= 2*F(240). F(2^e+1)=2*F(2^e1)+1 F(2^e2)=F(2^e3)+F(2^(e1)1) F(2^e+2)=F(2^e)+F(2^e1)+F(2^e2) I used the properties to test computed values. And for computing D.numbers.  Differences(n), Dif(n). SLOANE A132582. Dif(n)=F(n)F(n1). Dif(n) LT F(n1) Dif(n) as graph, or rule. At some n values, Dif(n) EQ F(m). n=(2^a1)*(2^b), m=(2^(a1)1)*(2^e) n=(2^a+1)*(2^b), m=(2^a1)*(2^e)  using 3value logic (0,1,2) i find: F(n)=F(n,2). F(2n,2) = F(n,3). D(7)=F(128,2)=F(64,3).  F(n fixed,s) give a polynomial n graded on s. I call them Pol. of Dedekind. Table F(n,s from 1 to ). 1. 1,2,3,... 2. 1,3,6,10,.. 3. 1,5,14,30,55,91,140,.. 4. 1,6,20,50,105,.. 6. 1,14,84,330,1001,.. 8. 1,20,168,.. First Dedekind polynomials: D1(x)=x D2(x)=(x)(x+1)/2 D3(x)=(x)(x+1)(2x+1)/6 D4(x)=(x1)(x)(x)(x+1)/12 A general expression is a openquestion.  Dedekind numbers as subserie of F(n), are computed on many ways. These methods, changing some parameters, gives a full set of series Dedekindlike. Fibo is one of them.  Another openquestion. All series with S(n) GT S(n1); S(n) LE 2*S(n1); Are Dedekindlike ??? 11Dic2018. JMMA 
20201007, 13:10  #4 
"Account archived."
Oct 2020
UE
31 Posts 
right place for Dedekind numbers?
Someone interested on the Dedekind numbers problem?
I computed M8 after 51' of Intel i3. I write a fast program. But computing serie A132581 is hard. I need bizarre computer power. Something like a Xeon 24Cores. Source is on github. 
20201007, 19:15  #5  
"Account archived."
Oct 2020
UE
31 Posts 
Oh, A132581...
Quote:
 Sequence OEIS A132581. "New values and one relationship." ONLY "with n multiple of 4"  New value obtained:* Discovered n F(n) Relationship    128 0002.4146.8204.0998; M7. 160 0001.0915.5249.3226.6679; Last known. 164* 0005.8783.4149.5816.1884; 168* 0014.1975.3406.9172.3630; 172* 0039.3953.4418.5632.4853; 176* 0051.2617.1120.4101.3308; 180* 0161.5425.7144.5831.0110; 184* 0225.4830.5719.5672.0898; 188* 0332.3658.7553.6968.2762; 192* 0337.9107.8507.6574.6508; F192+F252=F256. 196* 1925.8137.5568.0675.8426; 200* 5238.7009.8371.8298.2255; 204* 0001.6780.3451.1261.6143.1395; 224* 0022.8147.0929.0804.6885.3455; F224+F248=F256. 240* 0280.6521.8614.3437.7895.3894; 2*F240=F256. 248* 0538.4896.6299.6070.8905.4333; See 224. 252* 0561.2705.8120.8367.9216.1280; See 192. 253* 0561.3043.7223.8582.0165.4146; 254* 0561.3043.7226.2728.7586.6790; F256=F254+F128. 255* 0561.3043.7228.6875.5790.7787; F255+1=F256.  256 0561.3043.7228.6875.5790.7788; M8.  Comments All calculated with the x6486 OpenMP Windows10 EXE. Released on: https://github.com/JOSMANUE/DEDEKINDNUMBERS The program can calculate each F(n) until M8, with n multiple of 4, n between 8 and 256. Time spent is not O(n). Some numbers require computer power. Status. Pending: 208;212,216; 220,228; 232,236,244.  

20201007, 19:39  #6 
6809 > 6502
"""""""""""""""""""
Aug 2003
101×103 Posts
9,547 Posts 
Your crossposting in a new thread has been merged into this one. Please look for an existing thread before making a new one.

20201012, 22:41  #7 
"Daniel Jackson"
May 2011
14285714285714285714
11·59 Posts 
@JMARANDA: Your GitHub link is giving me a 404. Any chance you could fix that?

20201013, 08:21  #8 
"Oliver"
Sep 2017
Porta Westfalica, DE
2×13×19 Posts 
It's http://github.com/JOSMANUE/DedekindM851minutes., including the dot in the end.

20201015, 16:45  #9 
"Account archived."
Oct 2020
UE
31_{10} Posts 
Thanks for getting interested.
A cordial greeting. JMMA. https://github.com/JOSMANUE. https://github.com/JOSMANUE https://github.com/JOSMANUE/A132582Combinatory. https://github.com/JOSMANUE/DedekindM851minutes. m8 has a set approach. e.c. a combinatorial approach. Some points are missing on the d7d8 road. Even m8 requires in some cases quite computing power. The set approach is limited by memory. The combinatory is slower, but lacks limits. It works by calculating OEIS A132582 from every n. Several can be calculated in parallel. But it doesn't parallelize the individual calculation. In C the word limit is __int128, in C++ there is no word limit. In theory one could launch e.exe and reach d9. I'm updating github. If I get a parallel e.c. I'll put it on. Best regards. Well The Dedekind d9 is an unspotained peak. I encourage you to calculate it. There's nothing GIMPS can't do. 
20210318, 00:07  #10 
"Account archived."
Oct 2020
UE
31 Posts 
D(9) is really hard.
After D(8) , or F(256), few values of F() are known. I find some of these new values, like i post here time ago. I fail contact OEIS. Computing D(8) today requires some kind of fast Xeon cluster. And 4 hours time. I write one really more fastest program merging anything what i find and my best ideas. I post them on Github. It outperforms the previous program many times. Zero downloads, no forks. I contact previous team, at Nederlands: not more interested. Oh cruel world, give me one Cray and i move the D(9) !! 
20210318, 01:09  #11 
Aug 2006
13541_{8} Posts 

Thread Tools  
Similar Threads  
Thread  Thread Starter  Forum  Replies  Last Post 
GPU Computing newbie  lukerichards  GPU Computing  10  20190410 02:10 
History of Computing  Nick  Computer Science & Computational Number Theory  0  20171010 20:45 
GPU Computing Cheat Sheet (a.k.a. GPU Computing Guide)  Brain  GPU Computing  20  20151025 18:39 
Could a Distributed Computing approach help find the smallest Brier number?  jasong  Math  5  20070529 13:30 
The difference between P2P and distributed computing and grid computing  GP2  Lounge  2  20031203 14:13 