Predictions
- The first prediction in the table below is being made on Mon Apr 18 15:57:34 IST 2011. and makes use of the following asymptotic formula.
where we have fixed the multiplicative constant to give the exact answer for 62. We anticipate that the numbers will be of the correct order with the first three digits being correct.
- The second and third predictions are being made on Sun May 8 06:13:26 IST 2011 and makes use of a formula similar to $q_3(x)$ with more parameters added. The first parameter that is added in $r_3(x)$ is a shift i.e., $x=n+\textrm{shift}$. Thus $r_3(x)$ has two free parameters. The third prediction is based on adding a third parameter (to the shift and the overall scale) obtained by letting $c$ in the formula for $q_3(x)$ be a free parameter. This is represented by the function $s_3(x)$ — we now let $c$ in the formula for $q_3(x)$. These parameters were determined by fitting to the data in the range $[58,62]$. We expect that the predictions with larger number of parameters will provide better fits.
with $sh=-0.02844026154$ and
(3)with $sh=1.689159515$.
| $N$ | 61 | 62 | 63 | 64 |
|---|---|---|---|---|
| $q_3(N)$ | 17474590403967699 | used for fit | 46453074905306481 | 75522726337662733 |
| $r_3(N)$ | 17473162727829722 | 28517941261598224 | 46454377905586507 | 75528871134371025 |
| $s_3(N)$ | 17472949259247056 | 28518689103596432 | 46458481008470927 | 75541939909077818 |
| Exact | 17472947006257293 | 28518691093388854 | 46458506464748807 | 75542021868032878 |
| $N$ | 65 | 66 | 67 | 68 |
| $q_3(N)$ | 122556018966297693 | 198518226269824763 | 320988410810838956 | 518102330350099210 |
| $r_3(N)$ | 122572402629704284 | 198554960326767292 | 321063993755480353 | 518249988167731396 |
| $s_3(N)$ | 122606641782139492 | 198635634117555516 | 321241553598715473 | 518622446910206277 |
| Exact | 122606799866017598 | 198635761249922839 | 321241075686259326 | 518619444932991189 |
Results
Here are the results obtained on Wed Jun 1 13:01:58 IST 2011 by running a parallelized and modified version of Knuth's algorithm for about five months (360000 CPU hours). Below, c[N] denotes the number of solid partitions of N. So far we have added eighteen numbers (the last six are new), c[51] to c[68] to the numbers computed by Knuth and Mustonen-Rajesh. The number of solid partitions appear in the On-Line Encyclopedia of Integer Sequences as sequence A000293. Our list is given below:
c[1] = 1
c[2] = 4
c[3] = 10
c[4] = 26
c[5] = 59
c[6] = 140
c[7] = 307
c[8] = 684
c[9] = 1464
c[10] = 3122
c[11] = 6500
c[12] = 13426
c[13] = 27248
c[14] = 54804
c[15] = 108802
c[16] = 214071
c[17] = 416849
c[18] = 805124
c[19] = 1541637
c[20] = 2930329
c[21] = 5528733
c[22] = 10362312
c[23] = 19295226
c[24] = 35713454
c[25] = 65715094
c[26] = 120256653
c[27] = 218893580
c[28] = 396418699
c[29] = 714399381
c[30] = 1281403841
c[31] = 2287986987
c[32] = 4067428375
c[33] = 7200210523
c[34] = 12693890803
c[35] = 22290727268
c[36] = 38993410516
c[37] = 67959010130
c[38] = 118016656268
c[39] = 204233654229
c[40] = 352245710866
c[41] = 605538866862
c[42] = 1037668522922
c[43] = 1772700955975
c[44] = 3019333854177
c[45] = 5127694484375
c[46] = 8683676638832
c[47] = 14665233966068
c[48] = 24700752691832
c[49] = 41495176877972
c[50] = 69531305679518c[51] = 116221415325837
c[52] = 193794476658112
c[53] = 322382365507746
c[54] = 535056771014674
c[55] = 886033384475166
c[56] = 1464009339299229
c[57] = 2413804282801444
c[58] = 3971409682633930
c[59] = 6520649543912193
c[60] = 10684614225715559
c[61] = 17472947006257293
c[62] = 28518691093388854
c[63] = 46458506464748807
c[64] = 75542021868032878
c[65] = 122606799866017598
c[66] = 198635761249922839
c[67] = 321241075686259326
c[68] = 518619444932991189
Credits: B. Srivatsan (a sophomore at IIT Madras) for writing all the code which was parallelized by Suresh Govindarajan(SG). All runs were carried out by SG on machines at the High Performance Computing Environment at IIT Madras.
Disclaimer: While a lot of effort has gone into ensuring the correctness of our results, this has not been independently verified. Our results do agree with known results up to N=50 and could be considered as an independent verfication of the 2003 Mustonen-Rajesh result. as well as a validation of the numbers for N=51,52 that were generated using different code (v2.0); N=53-55 that were generated using v3.0 of the code and N=56-62 that were generated using v3.2 of the code.
Support the solid partitions project
How? Either run code or work on improving the code or help with setting up/running the portal. Contact solidpartitions at gmail.com
© 2011 Suresh Govindarajan