2
$\begingroup$

enter image description hereI want to know if there is a package in Mathematica that can give back some recursion given a list of polynomial in $f_{n}(h)$. For example I know there exists package RISC guess that where I give some list of rational numbers it can give me the recursion it satisfies. My list is as follows [1,0,-2,$-h^2+3$,$5h^2+1$,$-5h^2-11$,$-35h^2+15$,$-21h^4+140h^2+13$,$189h^4-84h^2-77$] I want a recursion of the from $$ \sum_{k=0}^{N}a_{k}(h)[n]f(n+k) =0 $$

where $a_{k}(h)[n]$ is a polynomial in $n$ over the rational field $Q(h)$.

An example done in Guess package is as follows enter image description here

[{1,1},{2,0},{3,-2}, {4,-h^2+3},{5,5*h^2+1},{5,5*h^2+1},{6,-5*h^2-11},{7,-35*h^2+15},{8,-21*h^4+140*h^2+13},{9,189*h^4-84*h^2-77},{10,-462*h^4-840*h^2+86},{11,-1848*h^4+2640*h^2+144},{12,-1485*h^6+15015*h^4-495*h^2-595},{13,19305*h^6-27027*h^4-16445*h^2+495},{14,-78507*h^6-102102*h^4+41041*h^2+1520},{15,-173745*h^6+636636*h^4+11375*h^2-4810},{16,-225225*h^8+2922920*h^6-816816*h^4-282100*h^2+2485},{17,3828825*h^8-9432280*h^6-4084080*h^4+559300*h^2+15675},{18,-22141548*h^8-15801500*h^6+19399380*h^4+474300*h^2-39560},{19,-15796638*h^8+218196836*h^6-15872220*h^4-4399260*h^2+6290},{20,-59520825*h^10+897498459*h^8-554268000*h^6-128741340*h^4+6807225*h^2+159105},{21,1249937325*h^10-4414328919*h^8-1127011600*h^6+476030940*h^4+11062275*h^2-324805},{22,-9445293585*h^10-903687785*h^8+10988363100*h^6-171609900*h^4-63677075*h^2-87075},{23,6418990305*h^10+101767530865*h^8-21565644100*h^6-3397876020*h^4+73363675*h^2+1592843}]

$\endgroup$
  • $\begingroup$ It would be useful to have an example that can be cut and pasted. $\endgroup$ – Daniel Lichtblau Apr 17 '18 at 14:44
  • $\begingroup$ I copy paste an example if it helps. It from RISC GUESS pacakge where they can do it for rational numbers but for me it's polynomials. $\endgroup$ – GGT Apr 17 '18 at 23:27
  • 1
    $\begingroup$ I cannot cut and paste an image... $\endgroup$ – Daniel Lichtblau Apr 18 '18 at 13:47
  • $\begingroup$ I see so which format is good for you ? I can send you a pdf file ? $\endgroup$ – GGT Apr 18 '18 at 23:39
  • $\begingroup$ Just put plain ascii text, in Mathematica input form, into the actual question. So one "row" could be the Mathematica list {10,-462*h^4-840*h^2+86} for example. $\endgroup$ – Daniel Lichtblau Apr 19 '18 at 14:15
1
$\begingroup$
pollist = {{1, 1}, {2, 0}, {3, -2}, {4, -h^2 + 3}, {5, 5*h^2 + 1}, {5,
     5*h^2 + 1}, {6, -5*h^2 - 11}, {7, -35*h^2 + 15}, {8, -21*h^4 + 
     140*h^2 + 13}, {9, 
    189*h^4 - 84*h^2 - 77}, {10, -462*h^4 - 840*h^2 + 
     86}, {11, -1848*h^4 + 2640*h^2 + 144}, {12, -1485*h^6 + 
     15015*h^4 - 495*h^2 - 595}, {13, 
    19305*h^6 - 27027*h^4 - 16445*h^2 + 495}, {14, -78507*h^6 - 
     102102*h^4 + 41041*h^2 + 1520}, {15, -173745*h^6 + 636636*h^4 + 
     11375*h^2 - 4810}, {16, -225225*h^8 + 2922920*h^6 - 816816*h^4 - 
     282100*h^2 + 2485}, {17, 
    3828825*h^8 - 9432280*h^6 - 4084080*h^4 + 559300*h^2 + 
     15675}, {18, -22141548*h^8 - 15801500*h^6 + 19399380*h^4 + 
     474300*h^2 - 39560}, {19, -15796638*h^8 + 218196836*h^6 - 
     15872220*h^4 - 4399260*h^2 + 6290}, {20, -59520825*h^10 + 
     897498459*h^8 - 554268000*h^6 - 128741340*h^4 + 6807225*h^2 + 
     159105}, {21, 
    1249937325*h^10 - 4414328919*h^8 - 1127011600*h^6 + 
     476030940*h^4 + 11062275*h^2 - 324805}, {22, -9445293585*h^10 - 
     903687785*h^8 + 10988363100*h^6 - 171609900*h^4 - 63677075*h^2 - 
     87075}, {23, 
    6418990305*h^10 + 101767530865*h^8 - 21565644100*h^6 - 
     3397876020*h^4 + 73363675*h^2 + 1592843}}[[All, 2]];

For free coefficients

coef0[n_] = FindSequenceFunction[Coefficient[#, h, 0] & /@ pollist, n]

gives

DifferenceRoot[{[FormalY],[FormalN]}[Function]{([FormalN]+4) (2936 [FormalN]^2-13255 [FormalN]+13103) [FormalY]([FormalN]+3)+(11818 [FormalN]^3-39180 [FormalN]^2+12312 [FormalN]+47530) [FormalY]([FormalN]+1)+(4441 [FormalN]^3-7769 [FormalN]^2-21002 [FormalN]+47530) [FormalY]([FormalN]+2)-5 ([FormalN]-5) [FormalN] (1431 [FormalN]-1199) [FormalY]([FormalN])==0,[FormalY](1)==1,[FormalY](2)==0,[FormalY](3)==-2}][n]

Check:

Table[coef0[k], {k, 1, 24}]

{1,0,-2,3,1,1,-11,15,13,-77,86,144,-595,495,1520,-4810,2485,15675,-39560,6290,159105,-324805,-87075,1592843}

For the coefficients at $h^2$

coef2[n_] = FindSequenceFunction[Coefficient[#, h, 2] & /@ pollist, n]

DifferenceRoot[{[FormalY],[FormalN]}[Function]{(550 [FormalN]^3-2640 [FormalN]^2+1220 [FormalN]+1110) [FormalY]([FormalN]+1)+(237 [FormalN]^3-1527 [FormalN]^2+2442 [FormalN]-672) [FormalY]([FormalN]+2)+25 ([FormalN]-5) [FormalN] (17 [FormalN]-5) [FormalY]([FormalN])+([FormalN]-2) ([FormalN]-1) (76 [FormalN]-297) [FormalY]([FormalN]+3)==0,[FormalY](1)==0,[FormalY](2)==0,[FormalY](3)==0,[FormalY](4)==-1,[FormalY](5)==5}][n]

etc.

BTW recurrences becomes simpler if to begin from the third line:

coef0[n_] = FindSequenceFunction[Coefficient[#, h, 0] & /@ pollist[[3 ;;]], n]

DifferenceRoot[{[FormalY],[FormalN]}[Function]{(5 [FormalN]+10) [FormalY]([FormalN])+(2 [FormalN]+7) [FormalY]([FormalN]+1)+([FormalN]+5) [FormalY]([FormalN]+2)==0,[FormalY](1)==-2,[FormalY](2)==3,[FormalY](3)==1,[FormalY](4)==1,[FormalY](5)==-11}][n]

$\endgroup$
  • $\begingroup$ Thanks for that. As I understand you gave the recursion for the constant term of the polynomials. What is Formal[Y], Formal[N] denotes? Is it possible to give a recursion for a bivariate table? The entry of the table would be $(d,coeff[f_d,h,2g])$. $f_d$ denote the $d-th$ polynomial in the list. $\endgroup$ – GGT Apr 21 '18 at 0:01
  • $\begingroup$ Yes, for the coefficients of $h^2$ too, changed the answer. As for Formal[Y] , Formal[N] it's mathematica's symbols introduced to not use y, n because those variables can be defined already. $\endgroup$ – Andrew Apr 21 '18 at 5:29
  • $\begingroup$ you mean you have a different recursion for coefficients of $h^2$ ? $\endgroup$ – GGT Apr 21 '18 at 7:18
  • $\begingroup$ To make sure the above recursion for the constant coefficient read as follows $ (5n+10)*y(n)+(2n+7)*y(n+1)+(2n+5)*y(n+2)=0$. $\endgroup$ – GGT Apr 21 '18 at 7:34
  • $\begingroup$ @GGT yes, for different coefficients recursions are different. $\endgroup$ – Andrew Apr 21 '18 at 7:44

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.