Classification of Partial Spread Functions in eight Variables
This page reports the numerical experiments related to the classification of the Partial Spread Boolean Functions in eight variables that I computed in 2008-2009. This work shows there are
extension classification stabilization
n time size time time class time psf
4 1 5 1 0 3 1 64374841666437120
5 15 233 55 10 22 10 20267057123180937216
6 69 4893 1162 385 341 6 1339989812392369324032
7 415 29691 7038 7246 3726 62 17833337132662061531136
8 1076 60943 14449 33501 9316 229 46056096661467073413120
9 681 31715 7516 8594 5442 19529 24520650576127040978944
10 219 8871 2109 698 1336 23 4731497045822911021056
11 75 2759 654 148 303 6 713809537614313684992
12 20 675 160 30 42 10 38019657690425327616
13 3 96 23 4 6 2 129740065512357888
14 0 11 3 0 1 59 44213490155520
15 0 3 1 0 1 11186 6579388416
16 0 2 0 0 1 0 200787
17 0 1 0 0 1 0 1
psf = (identifier) : (n-3) 16-bits numbers fix = (order of the fixator group). anf = (algebraic form) rnk = private spec= (spectrum)All the partial spreads of order n that we computed are extensions of the standard partial spread of order 3 : [ I: 0 ] [ 0 : I] [ I : I ]. The information on line psf= describes the generator matrices [ I : A ] of the other subspaces. A 16-bits number (A3 A2 A1 A0) represents a 4x4 matrix whose the rows are A0, A1, A2, A3.
psf-9.txt:rnk=999 psf-9.txt-fix=1008 psf-9.txt-anf=+25+135+45+345+126+36+136+46+246+346+27+127+137+47+147+247+347+18+28+38+138+348+358+268+368+278+378+478 psf-9.txt-spec= 136 [-16], 120 [16]and the other one is quadratic :
psf-9.txt-psf=5442 : 8137 27484 30180 33825 39544 57749 65213 psf-9.txt:rnk=999 psf-9.txt-fix=348364800 psf-9.txt-anf=+25+16+36+46+27+47+28+38 psf-9.txt-spec= 136 [-16], 120 [16],